CN111819853B

CN111819853B - Image block encoding device and image block encoding method

Info

Publication number: CN111819853B
Application number: CN201880090706.8A
Authority: CN
Inventors: 阿列克谢·康斯坦丁诺维奇·菲利波夫; 瓦西里·亚历斯维奇·拉夫特斯基
Original assignee: Huawei Technologies Co Ltd
Current assignee: Huawei Technologies Co Ltd
Priority date: 2018-03-07
Filing date: 2018-03-29
Publication date: 2023-07-18
Anticipated expiration: 2038-03-29
Also published as: EP3741123A1; JP7047119B2; CN111819852A; KR20200112964A; WO2019172802A1; US11856216B2; EP3738311A1; KR102419112B1; US11438618B2; WO2019172798A1; US20200404311A1; JP2021516016A; CN111819853A; CN111819852B; US20210014509A1

Abstract

Embodiments of symbol encoding and decoding of transform coefficients suitable for image and/or video encoding and decoding and the like are provided. Specifically, a plurality of the symbols are predicted, and only a prediction error signal is embedded in a code stream. The prediction error signal may have a distribution that may be efficiently encoded with CABAC or other variable length (entropy) encoding. Further, when an adaptive multi-core transform is used, a context in which the transform coefficient symbols are encoded by an entropy code is selected according to the transform used to obtain the transform coefficients.

Description

Image block encoding device and image block encoding method

The embodiment of the invention relates to the field of image processing such as still image and/or video image encoding and decoding.

Background

Video coding (video encoding and decoding) is widely used in digital video applications such as broadcast digital television, video transmission over the internet and mobile networks, real-time conversational applications for video chat, video conferencing, etc., DVD and blu-ray discs, video content acquisition and editing systems, and security applications for camcorders.

With the development of block-based hybrid video coding in the h.261 standard in 1990, new video coding techniques and tools have evolved and form the basis for new video coding standards. One of the goals of most video coding standards is to achieve a lower code rate than the previous standard, while guaranteeing image quality. Other video coding standards include MPEG-1 video, MPEG-2 video, ITU-T H.262/MPEG-2, ITU-T H.263, ITU-T H.264/MPEG-4 part 10, advanced video coding (Advanced Video Coding, AVC), ITU-T H.265, high efficiency video coding (High Efficiency Video Coding, HEVC), and extensions of these standards, such as scalability and/or three-dimensional (3D) extensions.

Disclosure of Invention

Embodiments of the invention are defined by the features of the independent claims and further advantageous implementations of the embodiments are defined by the features of the dependent claims.

According to some embodiments, different contexts are used for entropy encoding (encoding/decoding) of symbol prediction errors that are affected by horizontal and vertical transforms selected from a set of transforms provided by a multi-core transform technique. In particular, different transformations may be associated with respective different contexts.

According to one aspect, an image block encoding apparatus includes processing circuitry to: predicting (1320) a sign of a transform coefficient obtained by transforming (1300) the image block signal using one of a plurality of transforms; determining (1340) a symbol prediction error, the symbol prediction error indicating whether the symbol prediction is correct; selecting (1360) a context for entropy encoding the symbol prediction error according to one of the plurality of transforms; the symbol prediction error is encoded by applying entropy encoding using the selected context (1380).

According to one aspect, there is provided an apparatus for decoding an image block from a bitstream, the apparatus comprising processing circuitry for: predicting (1420) a symbol of transform coefficients to be inverse transformed using one of a plurality of transforms to obtain an image block signal; selecting (1440) a context for entropy decoding the symbol prediction error according to one of the plurality of transforms, the symbol prediction error indicating whether the symbol prediction is correct; decoding (1460) the symbol prediction error by applying entropy decoding using the selected context; the sign of the transform coefficient is determined (1480) from a predicted sign of the transform coefficient and a decoded sign prediction error of the transform coefficient.

Selecting an entropy coding context based on a transform for obtaining coefficients of a prediction symbol and a residual encoded with an entropy code may increase the efficiency of entropy coding, and thus also the efficiency of image coding applying symbol prediction.

In an exemplary implementation, the processing circuit is further to: the context is selected (1360) based on unsigned values of the transform coefficients.

Furthermore, since the symbol prediction efficiency may depend on the coefficient amplitude, providing different contexts according to the unsigned coefficient values may further improve the efficiency of entropy encoding and thus also the efficiency of image encoding.

Furthermore, the processing circuitry may be further configured to:

_– predicting (1320) a sign of each of M transform coefficients, M being an integer, greater than 1 and less than a number of transform coefficients obtained by transforming (1300) the image block;

_– determining (1340) a symbol prediction error for each of the M transform coefficients;

_– dividing the M transform coefficients into two lists through a threshold value, wherein a first list comprises transform coefficients with absolute values larger than the threshold value, and a second list comprises transform coefficients with absolute values equal to or smaller than the threshold value;

_– Selecting according to whether the transform coefficients of the M transform coefficients belong to the first list or the second list

(1360) A context for entropy encoding a symbol prediction error of a transform coefficient of the M transform coefficients;

_– the symbol prediction errors of the transform coefficients of the M transform coefficients are encoded by applying entropy encoding using the selected context (1380).

In an alternative implementation, the first list includes transform coefficients having absolute values equal to or greater than the threshold value, and the second list includes transform coefficients having absolute values less than the threshold value.

In particular, the complexity of predicting all coefficients will increase significantly due to hypothesis testing of sign bits. Complexity may be increased at both the encoding and decoding ends. Thus, complexity is reduced by predicting only M transform coefficient symbols. In some embodiments, predicting the sign of the M largest coefficients generally helps to predict the largest coefficients more reliably.

For example, the processing circuit is further configured to: the context is selected (1360) according to whether the prediction type of the image block is intra prediction or inter prediction. In another or alternative example, the processing circuit is further to: if the prediction type of the image block is intra prediction, the context is selected (1360) according to the intra prediction mode used to predict the image block. Specifically, the processing circuit is further configured to: predicting the image block using a DC intra prediction mode or a planar mode; the context is selected (1360) according to whether the image block is predicted using a DC intra prediction mode or a planar mode.

Since symbol prediction efficiency may depend on the size of the coefficients, the residuals obtained using intra-prediction (and specific intra-mode) and inter-prediction, respectively, may vary widely. Therefore, selecting a context according to the prediction type may further improve efficiency.

Specifically, the processing circuit is further configured to: predicting (1320) the sign of the transform coefficient according to a cost function, the cost function comprising estimating discontinuities along a boundary of the image block; the context is selected (1360) according to a number of block edges or a number of neighboring pixels that may be used to estimate a discontinuity on the boundary of the image block.

For example, the processing circuit is further configured to: the context is selected (1360) according to a ratio of edge pixels available for estimating discontinuities to a total number of pixels belonging to the image block. For example, the processing circuit is further configured to: the context is selected (1360) based on whether the image block is located at a left or upper boundary of a tile in which the image block is located.

The more adjacent pixels tested, the more accurate the prediction (in general), which can improve entropy coding efficiency.

In one exemplary implementation, the processing circuit is further to: the sign of the transform coefficients is predicted (1320) according to a cost function, wherein the cost function comprises transform differences between neighboring pixels of a neighboring transformed image block and predictions of the neighboring pixels calculated from a prediction signal of the image block.

The selection context according to the transform may be strongly correlated with the sign bit predictions calculated according to the cost function in the transform domain. In this case, the transformation has a direct effect on the distribution of the symbol prediction error.

According to one embodiment, each of the plurality of transforms is associated with one of a plurality of contexts, the processing circuitry further to: the context is selected (1360) from a plurality of contexts such that one of the plurality of transforms is associated with the selected context.

In one exemplary implementation, the processing circuit is further to: a context for entropy encoding the symbol prediction error is selected (1360) from one of the plurality of transforms each associated with one of a plurality of contexts, wherein a first transform of the plurality of transforms is associated with a same context of a second transform of the plurality of transforms if the probabilities of predicted symbol values of the first transform and the second transform differ by less than a predetermined value.

For example, the processing circuit is further configured to: predicting (1320) a sign of a transform coefficient obtained by transforming (1300) an image block signal using one of a plurality of transforms, wherein the transform (1300) comprises a horizontal transform and a vertical transform; the context is selected based on a combination of the horizontal transform and the vertical transform of the transform (1360).

In one example, each of the plurality of transforms is a combination of a horizontal transform and a vertical transform, the processing circuit further to: the context is selected (1360) such that the combination of the horizontal transform and the vertical transform is associated with the selected context.

Partitionable transforms are easier to implement and facilitate flexible selection of the appropriate transform.

For example, the number of mutually different contexts of the combination of the vertical transform and the horizontal transform is less than the number of the combination of the vertical transform and the horizontal transform, a first combination of vertical transform and horizontal transform being associated with the same context of a second combination of vertical transform and horizontal transform.

The processing circuit may also perform the entropy encoding using CABAC (1380).

According to one aspect, an image block encoding method includes the steps of: predicting (1320) a sign of a transform coefficient obtained by transforming (1300) the image block signal using one of a plurality of transforms; determining (1340) a symbol prediction error, the symbol prediction error indicating whether the symbol prediction is correct; selecting (1360) a context for entropy encoding the symbol prediction error according to one of the plurality of transforms; the symbol prediction error is encoded by applying entropy encoding using the selected context (1380).

According to one aspect, there is provided a method of decoding an image block from a bitstream, comprising the steps of: predicting (1420) a symbol of transform coefficients to be inverse transformed using one of a plurality of transforms to obtain an image block signal; selecting (1440) a context for entropy decoding the symbol prediction error according to one of the plurality of transforms, the symbol prediction error indicating whether the symbol prediction is correct; decoding (1460) the symbol prediction error by applying entropy decoding using the selected context; the sign of the transform coefficient is determined (1480) from a predicted sign of the transform coefficient and a decoded sign prediction error of the transform coefficient.

In one exemplary implementation, the context is selected (in the encoder and/or decoder) based on the unsigned value of the coefficient.

Furthermore, the method (encoding and/or decoding) performs (in the encoder and/or decoder) symbol prediction on each of M transform coefficients, M being an integer, greater than 1 and less than the number of transform coefficients obtained by transforming (1300) the image block; determining (1340) a symbol prediction error for each of the M transform coefficients; dividing the M transform coefficients into two lists through a threshold value, wherein a first list comprises transform coefficients with absolute values larger than the threshold value, and a second list comprises transform coefficients with absolute values equal to or smaller than the threshold value; selecting (1360) a context for entropy encoding a symbol prediction error of a transform coefficient of the M transform coefficients according to whether the transform coefficient of the M transform coefficients belongs to the first list or the second list; the symbol prediction errors of the transform coefficients of the M transform coefficients are encoded by applying entropy encoding using the selected context (1380).

In an alternative implementation of encoding and decoding (and encoder and decoder), the first list includes transform coefficients having absolute values equal to or greater than the threshold value, and the second list includes transform coefficients having absolute values less than the threshold value. For example, the method (encoding and/or decoding) also selects (1360) the context depending on whether the prediction type of the image block is intra-prediction or inter-prediction. In another or alternative example, the processing circuit is further to: if the prediction type of the image block is intra prediction, the context is selected (1360) according to the intra prediction mode used to predict the image block. In particular, the method (encoding and/or decoding) may predict the image block using a DC intra prediction mode or a planar mode; the context is selected (1360) according to whether the image block is predicted using a DC intra prediction mode or a planar mode.

In particular, the method (encoding and/or decoding) also predicts (1320) the sign of the transform coefficient according to a cost function comprising estimating a discontinuity along a boundary of the image block; the context is selected (1360) according to a number of block edges or a number of neighboring pixels that may be used to estimate a discontinuity on the boundary of the image block.

For example, the method (encoding and/or decoding) may select (1360) the context based on a ratio of edge pixels available for estimating discontinuities to a total number of pixels belonging to the image block. For example, the method may select (1360) the context based on whether the image block is located at a left or upper boundary of a tile in which the image block is located.

In one exemplary implementation, the method (encoding and/or decoding) predicts (1320) a sign of a transform coefficient according to a cost function, wherein the cost function includes transform differences between neighboring pixels of a neighboring transformed image block and predictions of the neighboring pixels calculated from a prediction signal of the image block.

In one exemplary implementation, the method (encoding and/or decoding) includes: a context for entropy encoding the symbol prediction error is selected (1360) from one of the plurality of transforms each associated with one of a plurality of contexts, wherein a first transform of the plurality of transforms is associated with a same context of a second transform of the plurality of transforms if the probabilities of predicted symbol values of the first transform and the second transform differ by less than a predetermined value.

For example, the method (encoding and/or decoding) may predict (1320) a sign of a transform coefficient obtained by transforming (1300) an image block signal using one of a plurality of transforms, wherein the transform (1300) comprises a horizontal transform and a vertical transform; the context is selected based on a combination of the horizontal transform and the vertical transform of the transform (1360).

In one example, each of the plurality of transforms is a combination of a horizontal transform and a vertical transform, the method (encoding and/or decoding) further selecting (1360) the context such that the combination of the horizontal transform and the vertical transform is associated with the selected context.

The entropy coding (1380) may be context-adaptive binary arithmetic coding (CABAC).

According to one embodiment, a non-transitory computer readable medium is provided storing a program comprising instructions which, when executed on a processor, perform all the steps of the method referred to above.

The details of one or more embodiments are set forth in the accompanying drawings and the description below. Other features, objects, and advantages will be apparent from the description and drawings, and from the claims.

Drawings

Embodiments of the present invention will be described in detail below with reference to the attached drawing figures, wherein:

FIG. 1 is a block diagram of an example of a video encoder for implementing an embodiment of the present invention;

fig. 2 is a block diagram of an exemplary architecture of a video decoder for implementing an embodiment of the present invention;

fig. 3 is a block diagram of an example of a video coding system for implementing an embodiment of the present invention;

FIG. 4 is a schematic diagram of symbol prediction;

FIG. 5 is a schematic diagram comparing symbol encoding and symbol prediction in H.264/AVC;

FIG. 6 is a schematic diagram of a current block and surrounding neighboring pixels;

FIG. 7 is a schematic diagram comparing known symbol predictions calculated using cost functions in the pixel domain with embodiments calculated using cost functions in the transform domain;

FIG. 8 is a flow chart of a process of transform coefficients associated with symbol prediction and coding;

FIG. 9 is a block diagram of an exemplary modification of the encoder according to FIG. 1 to include symbol prediction in the transform domain;

FIG. 10 is a block diagram of an exemplary modification of the decoder according to FIG. 2 to include symbol prediction in the transform domain;

FIG. 11 is an exemplary lookup table for selecting a CABAC context to entropy encode/decode a symbol prediction error;

FIG. 12 is an exemplary lookup table that selects a CABAC context to entropy encode/decode a symbol prediction error if certain combined contexts are incorporated;

FIG. 13 is a flow chart of the encoding method of the present invention;

FIG. 14 is a flow chart of a decoding method of the present invention;

fig. 15 is an example of a basis function, i.e. different combinations of horizontal and vertical transforms for a 4 x 4 block.

Detailed Description

In the following description, reference is made to the accompanying drawings which form a part hereof, and in which is shown by way of illustration specific aspects in which embodiments of the invention may be practiced. It is to be understood that embodiments of the invention may be used in other respects and include structural or logical changes not depicted in the drawings. The following detailed description is, therefore, not to be taken in a limiting sense, and the scope of the present invention is defined by the appended claims.

It will be appreciated that what is relevant to the described method also applies to the device or system corresponding to the method for performing the method and vice versa. For example, if one or more particular method steps are described, the corresponding apparatus may comprise one or more units, e.g., functional units, for performing the described one or more method steps (e.g., one unit performing the one or more steps, or a plurality of units each performing one or more of the plurality of steps), even if such one or more units are not explicitly described or illustrated in the figures. On the other hand, for example, if a particular apparatus is described in terms of one or more units (e.g., functional units), the corresponding method may include one step to perform the function of the one or more units (e.g., one step to perform the function of the one or more units, or multiple steps each to perform the function of one or more units of the plurality), even if such one or more steps are not explicitly described or illustrated in the figures. Furthermore, it is to be understood that features of the various exemplary embodiments and/or aspects described herein may be combined with each other, unless specifically indicated otherwise.

Video coding generally refers to the processing of a sequence of images that make up a video or video sequence. In the field of video coding, the term "frame" or "image" may be used as synonym, rather than the term "picture". Video coding includes two parts, video coding and video decoding. Performing video encoding on the source side typically includes processing (e.g., by compression) the original video image to reduce the amount of data needed to represent the video image (for more efficient storage and/or transmission). Video decoding is performed on the destination side, typically involving inverse processing with respect to the encoder to reconstruct the video image. Embodiments that relate to "encoding" of video images (or, in general, images, as will be explained below) are understood to relate to "encoding" and "decoding" of video images. The combination of the encoding portion and the decoding portion is also called CODEC (coding and decoding).

In the case of lossless video coding, the original video image may be reconstructed, i.e., the reconstructed video image has the same quality as the original video image (assuming no transmission loss or other data loss during storage or transmission). In the case of lossy video coding, compression (e.g., by quantization) is further performed to reduce the amount of data representing video images that cannot be fully reconstructed at the decoder side, i.e., the quality of the reconstructed video images is lower or worse than the original video images.

Several video coding standards, beginning with h.261, all belong to a "lossy hybrid video codec" (i.e. 2D transform coding that combines spatial prediction and temporal prediction in the pixel domain and applies quantization in the transform domain). Each image of a video sequence is typically divided into a set of non-overlapping blocks, and encoding is typically performed in units of blocks. In other words, on the encoder side, video is typically processed, i.e. encoded in blocks (video blocks), for example by: generating a prediction block using spatial (intra-picture) prediction and temporal (inter-picture) prediction; subtracting the predicted block from the current block (current processing/to-be-processed block) to obtain a residual block; the residual block is transformed in the transform domain and quantized to reduce the amount of data to be transmitted (compressed), while on the decoder side, the inverse process with respect to the encoder is applied to the encoded block or compressed block to reconstruct the current block for representation. Furthermore, the processing loop of the encoder is identical to that of the decoder, such that both will generate identical predictions (e.g., intra-prediction and inter-prediction) and/or reconstruct to process (i.e., encode) the subsequent block.

Since video image processing (also referred to as moving image processing) and still image processing (the term "processing" includes encoding) have many of the same concepts and technologies or tools, hereinafter, the term "picture or image" is used to refer to video images and/or still images (still pictures) of a video sequence (as explained above) to avoid unnecessary repetition and distinction between video images and still images when not needed. If the description refers only to still images (still pictures or still images), the term "still images (still pictures or still images)" should be used.

In the following embodiments, the encoder 100, the decoder 200 and the encoding system 300 are described with reference to fig. 1 to 3, and then embodiments of the present invention are described in more detail with reference to fig. 4 to 9.

Fig. 3 is a schematic block diagram of an embodiment of an encoding system 300 (e.g., image encoding system 300). The encoding system 300 includes a source device 310 for providing encoded data 330, such as an encoded image 330, to a destination device 320 for decoding the encoded data 330.

The source device 310 includes the encoder 100 or the encoding unit 100 and may additionally (i.e., optionally) include an image source 312, a preprocessing unit 314 (e.g., image preprocessing unit 314), and a communication interface or communication unit 318.

The image source 312 may include or be any type of image capturing device, such as a computer graphics processor for capturing real world images, and/or any type of image generating device, such as for generating computer animated images, or any device for capturing and/or providing real world images, computer animated images (e.g., screen content, virtual Reality (VR) images, and/or any combination thereof (e.g., augmented reality (augmented reality, AR) images).

The (digital) image is or can be regarded as a two-dimensional array or matrix of pixels having intensity values. The pixels in the array may also be referred to as pixels (a short form of picture element). The number of pixel points in the horizontal and vertical directions (or axes) of the array or image defines the size and/or resolution of the image. To represent a color, three color components are typically used, i.e., the image may be represented as or include an array of three pixel points. In RBG format or color space, an image includes corresponding arrays of red, green, and blue pixels. However, in video coding, each pixel is typically represented in a luminance/chrominance format or in a color space, e.g., YCbCr, including a luminance component represented by Y (sometimes also represented by L) and two chrominance components represented by Cb, cr. The luminance component Y represents luminance or grayscale intensity (e.g., as in a grayscale image), and the two chrominance components Cb and Cr represent chrominance or color information components. Accordingly, the YCbCr format image includes a luminance pixel point array of luminance pixel point values (Y) and two chrominance pixel point arrays of chrominance values (Cb and Cr). An image in RGB format may be converted to YCbCr format and vice versa, a process also known as color conversion or conversion. If the image is monochromatic, the image may include only an array of luminance pixels.

For example, the image source 312 may be a camera, image memory, or the like for capturing images, including or storing previously captured or generated images, and/or any type of interface (internal or external) to acquire or receive images. For example, the camera may be a local or integrated camera integrated in the source device and the memory may be a local or integrated memory, e.g. integrated in the source device. For example, the interface may be an external interface that receives images from an external video source, such as an external image capturing device, such as a camera, an external memory, or an external image generating device, such as an external computer graphics processor, computer, or server. The interface may be any type of interface according to any proprietary or standardized interface protocol, e.g. a wired or wireless interface, an optical interface. The interface to acquire image data 313 may be the same as communication interface 318 or may be part of communication interface 318. The communication interface may be any interface such as ethernet, WLAN, bluetooth or LTE interface, and may also be any wired or wireless interface, such as a satellite or optical interface. The transmission may be a peer-to-peer transmission, a broadcast transmission, or a multicast transmission.

Unlike the preprocessing unit 314 and the processing performed by the preprocessing unit 314, the image or image data 313 may also be referred to as an original image or original image data 313.

The preprocessing unit 314 is configured to receive (raw) image data 313, and to preprocess the image data 313 to obtain a preprocessed image 315 or preprocessed image data 315. For example, preprocessing performed by the preprocessing unit 314 may include clipping, color format conversion (e.g., from RGB to YCbCr), color correction, or denoising.

Encoder 100 is operative to receive preprocessed image data 315 and provide encoded image data 171 (e.g., details are further described with respect to fig. 1).

The communication interface 318 of the source device 310 may be used to receive the encoded image data 171 and send it directly to another device (e.g., the destination device 320) or any other device for storage or direct reconstruction, or to process the encoded image data 171 before storing the encoded data 330 and/or sending the encoded data 330 to another device (e.g., the destination device 320) or any other device for decoding or storage, respectively.

The destination device 320 includes a decoder 200 or decoding unit 200 and may additionally (i.e., optionally) include a communication interface or communication unit 322, a post-processing unit 326, and a display device 328.

The communication interface 322 of the destination device 320 is for receiving encoded image data 171 or encoded data 330, for example, directly from the source device 310 or from any other source, for example, a memory such as an encoded image data memory.

Communication interface 318 and communication interface 322 may be used to transmit encoded image data 171 or encoded data 330 over a direct communication link (e.g., a direct wired or wireless connection) between source device 310 and destination device 320, respectively, or over any type of network (e.g., wired (e.g., fiber optic cable, power line, copper cable, coaxial cable, or based on any other medium) or wireless network, or any combination thereof), or any type of private and public network, or any combination thereof.

For example, communication interface 318 may be used to package encoded image data 171 into a suitable format (e.g., data packets) for transmission over a communication link or network and may also include data loss protection and data loss recovery.

The communication interface 322, which forms a corresponding part of the communication interface 318, may for example be used for unpacking encoded data 330 to obtain encoded image data 171, and may also be used for performing data loss protection and data loss recovery, including error concealment, for example.

Both communication interface 318 and communication interface 322 may be configured as unidirectional communication interfaces, or bi-directional communication interfaces, as indicated by the arrows encoding image data 330 from source device 310 to destination device 320 in fig. 3, and may be used to: such as sending and receiving messages, such as establishing a connection, acknowledging and/or retransmitting lost or delayed data, including image data, and exchanging any other information related to the communication link and/or data transmission (e.g., encoded image data transmission).

Decoder 200 is operative to receive encoded image data 171 and provide decoded image data 231 or decoded image 231 (e.g., details will be further described with respect to fig. 2).

For example, the post-processor 326 of the destination device 320 is configured to post-process the decoded image data 231 (e.g., the decoded image 231) to obtain post-processed image data 327 (e.g., the post-processed image 327). The post-processing performed by post-processing unit 326 may include, for example, color format conversion (e.g., from YCbCr to RGB), color correction, cropping or resampling, or any other processing, e.g., for preparing decoded image data 231 for display by display device 328.

For example, the display device 328 of the destination device 320 is configured to receive post-processing image data 327 to display an image to a user or viewer. The display device 328 may be or include any type of display for representing a reconstructed image, such as an integrated or external display, or a monitor. For example, the display may include a Cathode Ray Tube (CRT), a liquid crystal display (liquid crystal display, LCD), a plasma display, an organic light emitting diode (organic light emitting diode, OLED) display, or any other type of display including beamers, holograms, or 3D/VR glasses.

Although fig. 3 depicts source device 310 and destination device 320 as separate devices, embodiments of the devices may also include source device 310 and destination device 320 or both corresponding functions of source device 310 and corresponding functions of destination device 320. In these embodiments, the source device 310 or corresponding function and the destination device 320 or corresponding function may be implemented using the same hardware and/or software or by hardware and/or software alone or any combination thereof.

From the description, it will be apparent to the skilled person that the presence and (exact) division of the different units or functions in the source device 310 and/or the destination device 320 as shown in fig. 3 may vary depending on the actual device and application.

Thus, the source device 310 and the destination device 320 shown in fig. 3 are merely exemplary embodiments of the present invention, and embodiments of the present invention are not limited to those shown in fig. 3.

Source device 310 and destination device 320 may comprise any of a variety of devices, including any type of handheld or stationary device, such as a notebook or laptop computer, a cell phone, a smart phone, a tablet or tablet computer, a camera, a desktop computer, a set-top box, a television, a display device, a digital media player, a video game console, a video streaming device, a broadcast receiver device, and may not use or use any type of operating system.

Encoder and encoding method

Fig. 1 is a schematic/conceptual block diagram of an embodiment of an encoder 100 (e.g., an image encoder 100), wherein the encoder 100 includes an input 102, a residual calculation unit 104, a transform unit 106, a quantization unit 108, an inverse quantization unit 110 and an inverse transform unit 112, a reconstruction unit 114, a buffer 116, a loop filter 120, a decoded image buffer (decoded picture buffer, DPB) 130, a prediction unit 160 (including an inter estimation unit 142, an inter prediction unit 144, an intra estimation unit 152, an intra prediction unit 154, a mode selection unit 162), an entropy encoding unit 170, and an output 172. The video encoder 100 shown in fig. 1 may also be referred to as a hybrid video encoder or a video encoder according to a hybrid video codec.

For example, the residual calculation unit 104, the transformation unit 106, the quantization unit 108, and the entropy encoding unit 170 constitute a forward signal path of the encoder 100, while the inverse quantization unit 110, the inverse transformation unit 112, the reconstruction unit 114, the buffer 116, the loop filter 120, the decoded image buffer (decoded picture buffer, DPB) 130, the inter prediction unit 144, and the intra prediction unit 154 constitute an inverse signal path of the encoder, wherein the inverse signal path of the encoder corresponds to the signal path of the decoder (see decoder 200 in fig. 2).

For example, the encoder is arranged to receive the image 101 or an image block 103 of the image 101 via the input 102, for example, an image forming a video or a sequence of images of a video sequence. Image block 103 may also be referred to as a current image block or image block to be encoded, and image 101 may be referred to as a current image or image to be encoded (especially in video encoding, for example, to distinguish a current image from other images of a previously encoded and/or decoded image of the same video sequence (i.e., a video sequence that also includes the current image)).

Embodiments of encoder 100 may include a partitioning unit (not depicted in fig. 1), e.g., also referred to as an image partitioning unit, for partitioning image 103 into a plurality of blocks (e.g., blocks such as block 103), typically into a plurality of non-overlapping blocks. The segmentation unit may be used to use the same block size and corresponding grid defining the block size for all images of the video sequence or to change the block size between images or subsets or groups of images and to hierarchically segment each image into corresponding blocks. The term "block" refers to a rectangular (not necessarily square, but possibly square) portion of an image.

As with image 101, block 103 is or may be a two-dimensional array or matrix of pixels having intensity values (pixel values), although the dimensions of image 101 are smaller. In other words, for example, block 103 may include one pixel array (e.g., one luminance array in the case of a monochrome image 101) or three pixel arrays (e.g., one luminance array and two chrominance arrays in the case of a color image 101) or any other number and/or type of arrays depending on the color format used. The number of pixels in the horizontal and vertical directions (or axes) of block 103 defines the size of block 103.

The encoder 100 as shown in fig. 1 is used to encode the image 101 block by block, for example, performing encoding and prediction for each block 103.

The residual calculation unit 104 is configured to calculate the residual block 105 pixel by pixel (pixel by pixel) from the image block 103 and the prediction block 165, for example, by subtracting pixel values of the prediction block 165 (further details regarding the prediction block 165 will be provided later) from pixel values of the image block 103, to obtain the residual block 105 in the pixel domain.

The transform unit 106 is configured to apply a transform, such as a spatial frequency transform or a linear spatial (frequency) transform, such as a discrete cosine transform (discrete cosine transform, DCT) or a discrete sine transform (discrete sine transform, DST), to pixel values of the residual block 105 in the transform domain, obtaining transform coefficients 107. Transform coefficients 107, which may also be referred to as transform residual coefficients, represent residual block 105 in the transform domain.

Transform unit 106 may be used to apply integer approximations of DCT/DST (e.g., core transforms specified for HEVC/H.265). Such integer approximations are typically scaled by a factor compared to the orthogonal DCT transform. In order to maintain the norms of the residual block of the forward inverse transform process, other scaling factors are applied during the transform process. The scaling factor is typically chosen according to some constraint, e.g., the scaling factor is a power of 2 for the shift operation, a trade-off between bit depth of the transform coefficients, accuracy, and implementation cost, etc. For example, a specific scaling factor is specified for the inverse transform by the inverse transform unit 212 on the decoder 200 side (and a corresponding inverse transform by the inverse transform unit 112 on the encoder 100 side, for example), and a corresponding scaling factor may be specified for the forward transform by the transform unit 106 on the encoder 100 side, for example, accordingly.

The quantization unit 108 is configured to quantize the transform coefficient 107 to obtain a quantized coefficient 109, for example, by applying scalar quantization or vector quantization. The quantized coefficients 109 may also be referred to as quantized residual coefficients 109. For example, for scalar quantization, different scaling may be applied to achieve finer or coarser quantization. A smaller quantization step corresponds to finer quantization, while a larger quantization step corresponds to coarser quantization. The applicable quantization step size may be indicated by a quantization parameter (quantization parameter, QP). The quantization parameter may be, for example, an index of a predefined set of applicable quantization step sizes. For example, a smaller quantization parameter may correspond to fine quantization (smaller quantization step size) and a larger quantization parameter may correspond to coarse quantization (larger quantization step size) and vice versa. Quantization may comprise dividing by a quantization step, while corresponding or inverse dequantization performed by the inverse quantization unit 110 or the like may comprise multiplying by the quantization step. Embodiments according to HEVC may be used to determine quantization step sizes using quantization parameters. In general, the quantization step size may be calculated from quantization parameters using fixed-point approximations of equations including division. Quantization and dequantization may introduce other scaling factors to recover the norm of the residual block, which may be modified due to scaling used in the fixed-point approximation of the equation of the quantization step size and quantization parameters. In one exemplary implementation, the inverse transform and the dequantized scaling may be combined. Alternatively, custom quantization tables may be used and signaled from the encoder to the decoder in the code stream. Quantization is a lossy operation, with the loss increasing with increasing quantization step size.

For example, an embodiment of the encoder 100 (or the quantization unit 108) may be used to output a quantization scheme and a quantization step size by corresponding quantization parameters so that the decoder 200 may receive and apply corresponding inverse quantization. For example, embodiments of encoder 100 (or quantization unit 108) may be used to output quantization schemes and quantization step sizes directly or by entropy encoding unit 170 or any other entropy encoding unit.

For example, the inverse quantization unit 110 is configured to apply inverse quantization of the quantization unit 108 to quantized coefficients to obtain dequantized coefficients 111 by applying an inverse of a quantization scheme applied by the quantization unit 108 according to or using the same quantization step as the quantization unit 108. The dequantized coefficients 111 may also be referred to as dequantized residual coefficients 111, corresponding to the transform coefficients 108, but the dequantized coefficients 111 are typically different from the transform coefficients due to quantization-induced losses.

The inverse transform unit 112 is for applying an inverse transform of the transform applied by the transform unit 106, for example, an inverse discrete cosine transform (discrete cosine transform, DCT) or an inverse discrete sine transform (discrete sine transform, DST), to obtain an inverse transform block 113 in the pixel domain. The inverse transform block 113 may also be referred to as an inverse transformed dequantized block 113 or an inverse transformed residual block 113.

The reconstruction unit 114 is configured to combine (e.g., add) the inverse transform block 113 and the prediction block 165 by pixel-by-pixel addition decoding the pixel values of the residual block 113 and the pixel values of the prediction block 165, and the like, to obtain a reconstructed block 115 in the pixel domain.

A buffer unit 116 (or simply "buffer" 116), e.g., a column buffer 116, is used to buffer or store reconstructed blocks and corresponding pixel values, e.g., for intra estimation and/or intra prediction. In other embodiments, the encoder may be configured to use unfiltered reconstructed blocks and/or corresponding pixel values stored in the buffer unit 116 for any type of estimation and/or prediction.

For example, the loop filtering unit 120 (or simply "loop filter" 120) is configured to filter the reconstruction block 115 by applying a sample-adaptive offset (SAO) filter or other filter such as a sharpening filter, a smoothing filter, or a collaborative filter to obtain a filter block 121. The filtering block 121 may also be referred to as a filtering reconstruction block 121. Other or additional filters may be applied in the loop.

An embodiment of the loop filter unit 120 may comprise (not shown in fig. 1) a filter analysis unit and an actual filter unit, wherein the filter analysis unit is arranged to determine loop filter parameters for the actual filter. The filter analysis unit may be adapted to apply fixed predetermined filter parameters to the actual loop filter, to adaptively select filter parameters from a predetermined set of filter parameters, or to adaptively calculate filter parameters for the actual loop filter.

Embodiments of the loop filter unit 120 may comprise (not shown in fig. 1) one or more filters (loop filter components/sub-filters), e.g. one or more different kinds or types of filters connected in series or in parallel or in any combination, wherein each filter may comprise a filter analysis unit alone or in combination with other filters of the plurality of filters to determine the respective loop filter parameters, as described in the preceding paragraphs. For example, embodiments of encoder 100 (loop filter unit 120, respectively) may be used to entropy encode output loop filter parameters directly or through entropy encoding unit 170 or any other entropy encoding unit so that, for example, decoder 200 may receive and apply the same loop filter parameters for decoding.

The decoded picture buffer (decoded picture buffer, DPB) 130 is used to receive and store the filter block 121. The decoded image buffer 130 may also be used to store other previously filtered blocks, such as the previously reconstructed and filtered block 121 of the same current image or a different image (e.g., a previously reconstructed image), and to provide a complete previously reconstructed (i.e., decoded) image (and corresponding reference blocks and pixels) and/or a partially reconstructed current image (and corresponding reference blocks and pixels) for inter estimation and/or inter prediction.

Other embodiments of the present invention may also be used to perform any type of estimation or prediction, e.g., intra/inter estimation and intra/inter prediction, using the previously filtered block and corresponding filtered pixel values of decoded image buffer 130.

Prediction unit 160, also referred to as block prediction unit 160, is configured to receive or obtain image block 103 (current image block 103 of current image 101) and decoded or at least reconstructed image data, e.g., reference pixels of the same (current) image from buffer 116 and/or decoded image data 131 from one or more previously decoded images of decoded image buffer 130, and to process such data for prediction, i.e., to provide prediction block 165, which prediction block 165 may be inter prediction block 145 or intra prediction block 155.

The mode selection unit 162 may be used to select a prediction mode (e.g., intra-prediction or inter-prediction mode) and/or a corresponding prediction block 145 or 155 for use as the prediction block 165 to calculate the residual block 105 and reconstruct the reconstructed block 115.

Embodiments of mode selection unit 162 may be used to select a prediction mode (e.g., from those supported by prediction unit 160) that provides a best match or, in other words, a minimum residual (minimum residual refers to better compression for transmission or storage), or a minimum signaling overhead (minimum signaling overhead refers to better compression for transmission or storage), or both. The mode selection unit 162 may be used to determine a prediction mode from the rate-distortion optimization (rate distortion optimization, RDO), i.e. to select the prediction mode that provides the least rate-distortion optimization, or to select the prediction mode of the associated rate-distortion that at least meets the prediction mode selection criteria.

Hereinafter, the prediction process (e.g., the prediction unit 160) and the mode selection (e.g., by the mode selection unit 162) performed by the exemplary encoder 100 will be explained in more detail.

As described above, the encoder 100 is configured to determine or select a best or best prediction mode from a (predetermined) set of prediction modes. For example, the set of prediction modes may include intra prediction modes and/or inter prediction modes.

The set of intra prediction modes may include 32 different intra prediction modes, e.g., non-directional modes such as DC (or mean) mode and planar mode, or directional modes such as defined in h.264; or may include 65 different intra-prediction modes, e.g., a non-directional mode such as a DC (or mean) mode and a planar mode, or a directional mode as defined in h.265.

The set of inter prediction modes (or possible inter prediction modes) depends on the available reference picture (i.e., the previously at least partially decoded picture as stored in the DBP 230) and other inter prediction parameters, e.g., whether the reference picture is an entire reference picture or only a portion of the reference picture, e.g., a search window region of a region surrounding the current block, for searching for a best matching reference block; and/or whether pixel interpolation, e.g., half-pixel and/or quarter-pixel interpolation, is applied.

In addition to the above prediction modes, a skip mode and/or a direct mode may be applied.

Prediction unit 160 may also be used to partition block 103 into smaller block partitions or sub-blocks, for example, iteratively using quad-tree-partition (QT), binary-tree-partition (BT), or trigeminal-tree-partition (TT), or any combination thereof; for example, each of the block partitions or sub-blocks is predicted, wherein the mode selection includes: a tree structure of the partition block 103 is selected and a prediction mode is applied to each of the block partitions or sub-blocks.

The inter estimation unit 142, also referred to as inter image estimation unit 142, is configured to receive or acquire the image block 103 (the current image block 103 of the current image 101) and the decoded image 131, or at least one or more previous reconstructed blocks, for example, one or more reconstructed blocks of other/different previous decoded images 131, and perform inter estimation (inter estimation or inter picture estimation). For example, the video sequence may comprise the current image and the previous decoded image 131, or in other words, the current image and the previous decoded image 131 may be part of or constitute a sequence of images of the video sequence.

For example, the encoder 100 may be configured to select a reference block from a plurality of reference blocks of the same or different images of a plurality of other images, and provide the reference image (or reference image index … …) and/or an offset (spatial offset) between a position (x, y coordinates) of the reference block and a position of the current block as the inter estimation parameter 143 to the inter prediction unit 144. This offset is also called Motion Vector (MV). Inter-frame estimation is also referred to as motion estimation (motion estimation, ME), and inter-frame prediction is also referred to as motion prediction (motion prediction, MP).

The inter prediction unit 144 is configured to acquire (e.g., receive) inter prediction parameters 143, and perform inter prediction according to or using the inter prediction parameters 143 to acquire an inter prediction block 145.

Although fig. 1 shows two different units (or steps) for inter coding, namely inter estimation 142 and inter prediction 144, both functions may be performed as one unit (or step) (inter estimation requires/includes calculating inter prediction blocks, i.e. the or "one" type of inter prediction 144), e.g. by iteratively testing all possible or predetermined subsets of possible inter prediction modes while storing the current best inter prediction mode and the corresponding inter prediction block and taking the current best inter prediction mode and the corresponding inter prediction block as (final) inter prediction parameters 143 and inter prediction blocks 145 without performing inter prediction 144 at other times.

The intra-frame estimation unit 152 is configured to obtain (e.g., receive) the image block 103 (current image block) and one or more previous reconstructed blocks, e.g., reconstructed neighboring blocks of the same image, for intra-frame estimation. For example, the encoder 100 may be configured to select an intra prediction mode from a plurality of (predetermined) intra prediction modes and provide it as an intra estimation parameter 153 to the intra prediction unit 154.

Embodiments of encoder 100 may be used to select an intra prediction mode based on optimization criteria, such as minimum residual (e.g., intra prediction mode that provides a prediction block 155 most similar to current image block 103) or minimum rate distortion, etc.

The intra prediction unit 154 is configured to determine an intra prediction block 155 according to intra prediction parameters 153, e.g. a selected intra prediction mode 153.

Although fig. 1 shows two different units (or steps) for intra coding, namely intra estimation 152 and intra prediction 154, both functions may be performed as one unit (or step) (intra estimation (typically/always) requires/includes calculating intra prediction blocks, i.e. the or "one" intra prediction 154), e.g. by iteratively testing all possible or predetermined subsets of possible intra prediction modes while storing the current best intra prediction mode and the corresponding intra prediction block and taking the current best intra prediction mode and the corresponding intra prediction block as (final) intra prediction parameters 153 and intra prediction blocks 155, without performing intra prediction 154 at other times.

The entropy encoding unit 170 is configured to apply an entropy encoding algorithm or scheme (e.g., a variable length coding (variable length coding, VLC) scheme, a context adaptive VLC (CALVC) scheme, an arithmetic coding scheme, a context adaptive binary arithmetic coding (context adaptive binary arithmetic coding, CABAC)) to the quantized residual coefficients 109, the inter-prediction parameters 143, the intra-prediction parameters 153, and/or the loop filter parameters, alone or in combination (or not at all), to obtain encoded image data 171 that may be output by the output 172, e.g., in the form of an encoded bitstream 171.

Fig. 2 illustrates an exemplary video decoder 200 for receiving encoded image data (e.g., encoded code stream) 171, for example, encoded by encoder 100, to obtain decoded image 231.

Decoder 200 includes an input 202, an entropy decoding unit 204, an inverse quantization unit 210, an inverse transformation unit 212, a reconstruction unit 214, a buffer 216, a loop filter 220, a decoded image buffer 230, a prediction unit 260, an inter prediction unit 244, an intra prediction unit 254, a mode selection unit 262, and an output 232.

The entropy decoding unit 204 is configured to perform entropy decoding on the encoded image data 171 to obtain quantized coefficients 209 and/or decoded encoding parameters (not shown in fig. 2), e.g., any or all of (decoded) inter-prediction parameters 143, intra-prediction parameters 153, and/or loop filter parameters.

In an embodiment of the decoder 200, the inverse quantization unit 210, the inverse transformation unit 212, the reconstruction unit 214, the buffer 216, the loop filter 220, the decoded image buffer 230, the prediction unit 260 and the mode selection unit 262 are configured to perform inverse processing of the encoder 100 (and its respective functional units) to decode the encoded image data 171.

In particular, the function of the inverse quantization unit 210 may be the same as that of the inverse quantization unit 110, the function of the inverse transformation unit 212 may be the same as that of the inverse transformation unit 112, the function of the reconstruction unit 214 may be the same as that of the reconstruction unit 114, the function of the buffer 216 may be the same as that of the buffer 116, the function of the loop filter 220 may be the same as that of the loop filter 120 (regarding an actual loop filter, since the loop filter 220 typically does not include a filter analysis unit that determines filter parameters from the original image 101 or the block 103, but receives (explicitly or implicitly) or acquires filter parameters for encoding from, for example, the entropy decoding unit 204), and the function of the decoded image buffer 230 may be the same as that of the decoded image buffer 130.

The prediction unit 260 may include an inter prediction unit 244 and an intra prediction unit 254, wherein the inter prediction unit 244 may have the same function as the inter prediction unit 144 and the intra prediction unit 254 may have the same function as the intra prediction unit 154. The prediction unit 260 and the mode selection unit 262 are typically used for performing block prediction and/or obtaining only the prediction block 265 from the encoded data 171 (without any other information of the original image 101) as well as receiving or obtaining (explicitly or implicitly) the prediction parameters 143 or 153 and/or information about the selected prediction mode from, for example, the entropy decoding unit 204.

For example, decoder 200 is configured to output decoded image 231 via output 232, for presentation to a user or for viewing by a user.

The video codec may include techniques for multi-core transform coding, i.e., techniques capable of selecting a transform core (basis function) from multiple transform cores, as described in section 2.5.1 of the "joint development test model 1 algorithm description (Algorithm Description of Joint Exploration Test Model 1)" of the second conference of ISO/IEC JTC 1/SC 29/WG 11, jfet-B0021, by chen et al, ITU-T SG 16wp 3, san diego, usa, 20 to 26, and residual symbol prediction (residual sign prediction, RSP), further described below. Referring to fig. 1 and 2 described above, the multi-core transform may be implemented in the transform unit 106 and the inverse transform unit 110 of the encoder 100 and the inverse transform unit 212 of the decoder 200, respectively. One of the techniques is an adaptive multi-core transform (also denoted AMT). In addition to DCT-II and 4X 4DST-VII used in HEVC, the AMT scheme is used for residual coding of inter-coded blocks and intra-coded blocks. In addition to the current transform in HEVC, multiple selected transforms from the DCT/DST series are utilized. The newly introduced transformation matrices are DST-VII, DCT-VIII, DST-I and DCT-V.

The AMT applies to CUs less than 64×64, and whether the AMT applies to control by CU-level flags of all Transform Units (TUs) within the CU. When the CU level flag is equal to 0, DCT-II is applied in the CU to encode the residual. For each TU within the AMT-enabled CU, two other flags are sent to identify the horizontal and vertical transforms to be used.

For intra residual coding, a mode dependent transform candidate selection procedure is used, since the residual statistics for different intra prediction modes are different. As shown in table 1, three transform subsets are defined, and the transform subsets are selected according to intra prediction modes, as shown in table 2.

Table 1: three predefined transformation candidate sets

Using the subset concept, when the CU level flag is equal to 1, the transform subset is first identified using the intra prediction mode for each TU according to table 1. Thereafter, for each of the horizontal and vertical transforms (denoted as H and V in table 2), one of the two transform candidates in the identified subset of transforms in table 2 is selected, for example, according to the flag of the explicit indication. It should be noted that this description relates to AMT processes.

In general, the present invention is applicable to any other way of indicating a transform that obtains coefficients. For example, table 1 may include more or less different transforms and sets of transforms. The signaling using table 2 may also be replaced with other types of signaling.

Table 2: selected (H) horizontal transform set and (V) vertical transform set for each intra prediction mode

However, for inter prediction residuals, all inter modes as well as horizontal and vertical transforms use only one transform set consisting of DST-VII and DCT-VIII.

Residual symbol prediction may be performed in the transform domain. If the codec uses multiple core transforms (e.g., in the case of AMT), the codec operates using residual signals in multiple transform domains. The prediction of the transform coefficient symbols should take into account the transform domain to which the transform coefficients belong. Since the probability of prediction error depends on the core transform used, the coding efficiency of transform domain residual symbol prediction (transform domain residual sign prediction, TD-RSP) is adversely affected if there is no coordination between the core transform selection and the symbol prediction.

As described above, some image and video codecs encode quantized transform coefficients. Non-zero transform coefficients are signed, i.e. consist of an absolute value and a positive or negative sign. Encoding the sign of a coefficient requires a bit indicating whether it is positive or negative. In other words, a sign bit value of 0 may indicate a positive sign and a sign bit value of 1 may indicate a negative sign, or vice versa.

The most advanced video coding standards do not use entropy coding of quantized transform coefficient symbols. In h.264/AVC and h.265/HEVC, the symbol data is considered to be equiprobable (probability of occurrence of sign etc.), so it is encoded in CABAC bypass mode. However, the symbol data may be predicted by analyzing discontinuities between reconstructed blocks. The probability of the occurrence of the sign of the quantized transform coefficient providing the smaller discontinuity is higher than the probability of the occurrence of the sign of the quantized transform coefficient improving the discontinuity. Several methods are based on such statistical properties, such as the contribution JCTVC-a115"Video coding technology proposal by Fujitsu (hereinafter JCTVC-a 115) and US 2017/0142444 A1 (hereinafter US' 444), both of which are incorporated herein by reference, at the first JCT-VC conference of dellesden, germany, kazui et al, 2010.

The technique of JCTCVC-A115 is shown in FIG. 4. The technique estimates the sign of the transform coefficients of the current block 410 from pixels in neighboring blocks (previously processed (i.e., encoded or decoded), in this example block, top, upper left corner, and left neighboring blocks of the current block 410) and encodes the difference of the estimated sign from the original sign with CABAC (0: the same, 1: different). If the symbol is well estimated (predicted), the difference tends to zero and CABAC can improve coding efficiency.

In general, pixel a at the boundary of the current block 410 is highly correlated with pixels at the same boundary on one side of the neighboring block. This attribute is used to predict the sign of the transform coefficients in the current block 410. Assuming there are M non-zero coefficient (C (1), C (2) … … C (M)) symbols, the symbols of these coefficients will be predicted in the current block 410, possible combinations of these symbols K (S) _K (1)、S _K (2)……S _K (M)) is 2 ^M . The combination ranges from (+, + … … +) to (-, - … … -). In this example, the transform coefficients are discrete cosine transform (discrete cosine transform, DCT) coefficients. The determined coefficients include their absolute rank values (unsigned values, i.e., magnitudes) and the sign combination K. These coefficients are inverse transformed into the pixel domain and inverse quantized (i.e., scaled and rounded) to produce a block of quantized residuals 440 (prediction error). The prediction error 440 in the pixel domain is added to the prediction block 430 to obtain the reconstructed block 420. Will be regarded asThe reconstructed boundary pixels B at the upper and left boundaries of the front block 410 are compared to the pixels a extrapolated from the neighboring blocks (indicated by the arrows in fig. 4). This is performed for all combinations K. A combination of symbols K is defined as the estimated symbol, which minimizes the square error between pixel a and pixel B. The sum of absolute differences can also be minimized.

Fig. 5 shows a comparison of the method proposed in JCTVC-a115 with conventional h.264/AVC symbol encoding, and table 3 below summarizes certain steps performed by the method of JCTVC-a 115.

Table 3: method steps for coefficient symbol prediction and encoding

As can be seen from the top part of fig. 5, the conventional h.264/AVC method encodes all symbols by CABAC bypass coding.

Fig. 6 shows a detailed view of current block 610 and pixel 630 on top of current block 610 and pixel 620 to the left of current block 610. Pixels 630 and 620 belong to neighboring blocks.

In the spatial (pixel) domain, the cost function F is defined as follows:

where N and M are the height and width of the block, respectively. As can be seen from equation 1, if the pixel value Y _i,0 And Y _0,j The value of (i= … … N and j= … … M) is equal to the value of two rows (Z _1,j And Z _2,j ) And two columns (X) _i,-2 And X _i,-1 ) The value of the cost function F is smaller if the pixel values of the pixels are similar.

The technique presented in the above summary document involves performing an inverse transformation of the symbol estimates, since the difference in pixel values is used to calculate the cost function value F for a given set of symbols, which is determined by the hypothesis being examined (which hypothesis corresponds to the particular combination K). Although fast estimation methods exist, the computational cost of the conversion to spatial domain is still high, which is a major drawback of these methods.

In view of this, the present invention provides embodiments that can reduce the number of operations required to perform symbol prediction. This may be achieved by calculating the cost estimate in the transform domain instead of in the spatial domain.

Specifically, according to one embodiment, there is provided an image block decoding apparatus including processing circuitry for: predicting a sign of a plurality of coefficients of a transformed image block according to a cost function, wherein the cost function comprises a transformed difference between neighboring pixels adjacent to the transformed image block and predictions of the neighboring pixels calculated from a prediction signal of the image block; reconstructing the sign of the plurality of coefficients from the predicted sign.

It should be noted that the plurality of coefficients may be a finite M non-zero coefficients. As described above, the M coefficients may be M coefficients having the largest magnitudes among the coefficients of the current block. However, the present invention is not limited thereto, and the M coefficients may be only the first M non-zero coefficients in a predetermined order in the current block. The predetermined order may be a scanning order, wherein the scanning order may be predetermined by defining the scanning order in a standard or by defining a set of scanning orders in a standard, the set of scanning orders may be configured by signaling within the code stream or implicitly derived from some different parameter in the code stream, such as a prediction mode, etc. Alternatively, the scan order may be sent entirely in the code stream. In the case of quantization applied to coefficients, the coefficients are typically already quantized coefficients.

Thus, in contrast to fig. 4 discussed above, the prediction signal 430 is not added to the reconstructed residual 440, but the prediction signal 430 is subtracted from the neighboring pixel values, because the inverse transform is not performed, and thus the reconstructed residual signal 440 is not used. Specifically, instead of performing inverse transformation on the residual signal to obtain a reconstructed boundary pixel, a difference between a pixel adjacent to the current block and a pixel of a prediction signal of the current block extrapolated to an area of the adjacent pixel is subjected to forward transformation.

Furthermore, according to an exemplary implementation, the cost function comprises a sum of squared transform differences between neighboring pixels of the neighboring transformed image block and predictions of the neighboring pixels calculated from the prediction signal of the image block.

The use of square norms has the advantage of pasmodic identity. According to the pasmodic identity, the sum of squared differences (sum of squared difference, SSD) calculated in the transform domain of orthogonal transforms (e.g., discrete cosine transform (discrete cosine transform, DCT) and discrete sine transform (discrete sine transform, DST)) should provide the same result as the SSD calculated in the spatial (pixel) domain. Thus, the coding efficiency of the proposed technique should not be lower than the efficiency of the techniques described above in connection with JCTVC-a113 and US' 444.

Since symbol prediction is performed in the transform domain, the computational complexity of the above-described embodiments can be greatly reduced since the computation required to perform the inverse transform is eliminated. Thus, implementations of the present embodiments may be hardware friendly and do not require additional RAM and ROM memory buffers. In addition, the forward transform module 106 for performing the transform may be reused. Only one boundary pixel need be fetched from the column buffer.

According to one embodiment, the symbol prediction error is also encoded using the CABAC context model, rather than using the equiprobable symbol value encoding (bypass). However, the symbol estimation procedure performed in a similar manner in the encoder and in the decoder is performed differently from the procedure described above in connection with fig. 4 and 5.

A schematic diagram of an exemplary implementation of the above-mentioned embodiment can be seen from fig. 7. The following are the particularities of this implementation:

_– performing symbol prediction in a transform domain;

_– instead of reconstructing the current block and subtracting its boundary pixels from the corresponding pixels of the neighboring block, the prediction signal is propagated to the neighboring block region.

For simplicity, only the columns of neighboring blocks are considered in fig. 7.

Depending on the framework to which the present embodiment is applied, the upper column, right row, and even the lower column may also be used. In other words, symbol prediction may use any available boundary pixels. In the above example, the available blocks are assumed to be the top block, the left block, and the upper left corner block. This corresponds to the following assumption: the processing order of the blocks is left to right, top to bottom, which is common in current codecs. In general, in a decoder, any previously decoded neighbor blocks (in decoding order) may be used. Accordingly, in the encoder, any previously encoded neighbor block may be used.

In one implementation of the embodiment, the cost function F is redefined to use the square of the pixel difference instead of the modulus (sum of absolute differences):

wherein N and M are the height and width of the block; x, Y and Z are still defined as described in connection with fig. 6.

The reconstruction block 710 is composed of a prediction 720 portion and a residual 730 portion:

Y _i,j ＝P _i,j +R _i,j ，

wherein P is _i,j Is the predicted pixel at position i, j, is R _i,j Prediction error pixels (residuals) at positions i, j.

In this case, the components in equation 1a may be rearranged as follows:

wherein:

T _n ＝[2X _n,-1 -X _n,-2 -P _n,0 ]，

V _m ＝[2Z _-1,m -Z _-2,m -P _0,m ]，

Q _n ＝R _n,0 ，O _m ＝R _0,m 。

according to the pasmodic identity, the function F of equation 1a can be rewritten in the transform domain as the following equation 2:

in equation 2 above:

t _n ＝Trans1D(T _n )，

q _n ＝Trans1D(Q _n )，

v _n ＝Trans1D(V _n )，

o _n ＝Trans1D(O _n )，

where Trans1D () is a one-dimensional orthogonal transform.

Thus, a cost function can be calculated so that the sign of the quantized transform coefficients can be predicted in the transform domain.

To determine q _n 、o _n Two-dimensional transform r with residual _n,m ＝Trans2D(R _n,m ) The relation between R is written in a general form _n,m And r _n,m Relationship between:

wherein,,is the transform core (basis function). For example, the basis functions of a two-dimensional DCT (DCT 2) are as follows:

further, the following equation 3 is defined in the pixel domain (Q _n ) And transform domain (q _k ) Residual of column of current block at boundary with neighboring block:

The first equation above is r _k,l Is transformed back to Q _n Individual pixels (by Q _n Definition of (d). The transform kernel W is here orthogonal, i.e. the forward transform (second equation) and the inverse transform (first equation) coincide. Does not need to be to Q _n Perform 1D transform (first equal sign): in practice, q _k Is r _k,l Is the inverse 1D transform (second equal sign).

O is o _n (column) and r _n,m In other words, there is a similar relationship. Specifically, q _k (line coefficients) correspond to W ^0,l r _k,l ，o _n (column coefficient) corresponds to W ^m,0 r _k,l . Accordingly, with q _k In contrast, the zero index simply changes the position of o.

Thus, the above cost function F in equation 2 can be completely calculated in the frequency domain. On this basis, predictions of the sign of the plurality of coefficients are calculated.

In the image block decoding apparatus, the processing circuit may be configured to: analyzing a symbol prediction error from the coded code stream; reconstructing the symbol includes adding the parsed symbol prediction error to the predicted symbol.

Fig. 7 shows a comparison of the method described above in connection with fig. 4. In particular, reference symbols 710, 720 and 730 correspond to the respective symbols 420, 430 and 440, respectively, and represent a reconstructed block, a prediction block and a residual block. As also shown in fig. 1, the residual block is obtained by: the K-th hypothesis of M symbols is used for the transform coefficients of block a (symbol combination), and then the coefficients with symbols are inverse transformed and dequantized (block iq+idct).

To avoid such inverse transformation and inverse quantization for each test hypothesis, fig. 7 illustrates a method of an embodiment of the present invention. Specifically, the coefficients of block a (corresponding to column q at the boundary of block a and neighboring block 740 _n ) And a conversion difference column B _col A comparison was made. The transformed difference column B is obtained by _col (corresponding to t _n ): the propagation prediction signal 760 is subtracted from the neighbor block column 770 to obtain a difference signal 750 in the pixel domain, which difference signal 750 is converted into a variance difference. The transformation being any orthogonal transformation, e.g. a transformationTo the spectral domain, for example DFT, FFT, DCT or DST or integer versions thereof. The difference signal 750 corresponds to T _n ＝[2X _n,-1 -X _n,-2 -P _n,0 ]Wherein the propagation prediction 770 from neighboring blocks is (2X _n,-1 -X _n,-2 ) The prediction 760 of block 790 is P _n,0 . Acquisition of pass X _n,-2 And X is _n,-1 The extrapolated portion of the slope determination between (i.e., two columns at the boundary of the neighboring block) serves as the propagation prediction 770.

The comparison between a and B in fig. 7 is then made in the transform domain by the following cost function:

as can be seen from the above cost function, it corresponds to the cost function F in equation 2 above, but is limited to its column part based on the left-hand neighbor block, omitting the row part based on the top neighbor block. This is just one example for simple explanation. Any neighbor may be used for comparison in a similar manner, as will be clear to those skilled in the art.

Generally, according to this embodiment, the processing circuit is further configured to predict a sign of a plurality of coefficients of the transformed image block, comprising:

-calculating a transform difference B between neighboring pixels 770 of the neighboring transformed image block and a prediction 760 of neighboring pixels calculated from a prediction signal P of the image block 790;

-calculating a cost function F, wherein the cost function comprises a sum of the transformed difference and a squared transformed difference between transformed image blocks a, wherein according to the assumption set k=1.. 2^M symbol S _K (i) Each hypothesis K of (i=0..m-1) reconstructs the transformed image block a;

selecting as prediction symbol the hypothesis K of the symbol minimizing the cost provided by the cost function F (corresponding predicted sign=argmin _K F)。

For example, by comparing the symbol S _K (i) Is added to the coefficient C (i) parsed from the encoded code stream to reconstruct the transformed image block. This method is used to verify the K hypothesisThe best hypothesis to minimize the cost function is found. This assumption becomes the prediction symbol for block a. Furthermore, to obtain the symbols of block a, the prediction symbols are then added to the symbol prediction errors, which may be decoded from the code stream using Context-adaptive binary arithmetic coding (CABAC).

As explained above, in one particular exemplary implementation, the cost function F is provided by:wherein t is _n ＝Trans1D(T _n )，q _n ＝Trans1D(Q _n )，v _n ＝Trans1D(V _n )，o _n ＝Trans1D(O _n ) The method comprises the steps of carrying out a first treatment on the surface of the Wherein, trans1D ()' is one-dimensional orthogonal transformation, T _n ＝[2X _n,-1 -X _n,-2 -P _n,0 ]，V _m ＝[2Z _-1,m -Z _-2,m -P _0,m ]，Q _n ＝R _n,0 ，O _m ＝R _0,m The method comprises the steps of carrying out a first treatment on the surface of the Wherein P is a prediction signal, X and Y are adjacent pixels, and N and M are the height and width of the block where the prediction symbol is located, respectively.

It should be noted that adjacent pixels may be located at the horizontal and vertical boundaries of the image block. The vertical boundary may be a left boundary or a right boundary and the horizontal boundary may be a top boundary. It is advantageous if the top and left (and/or right) neighboring blocks have been decoded when the current block was processed (decoded). However, the embodiments of the present invention are not limited thereto. In general, one or more neighboring blocks that have been decoded (and thus are available) when decoding the current block may be used to predict the symbol. The cost function F then includes a corresponding smoothness check of the current block pixels with respect to those boundary pixels from the respective available neighbor blocks.

Fig. 8 illustrates a coding flow diagram provided by one embodiment. Fig. 8 also includes context selection and entropy coding (see block 808) and embedding into the code stream (see block 811) within the residual symbol prediction process.

The context determination process 802 uses quantized transform coefficients 801 corresponding to the transform and quantized residual, wherein the context determination process 802 operates in a manner similar to that described in US' 444.

US'444 provides more details about symbol estimation. For example, it shows a different prediction symbol encoding method than step 3 of table 3 above. This modification is achieved by introducing two lists of predicted symbols (modification of step 1 of table 3). The prediction symbols belonging to the two lists are encoded using different CABAC contexts. The following rules are specified to populate these lists:

-the first list is greater than a predefined threshold T in magnitude ₁ Is used for filling the coefficient symbols of the (a). The total number of symbols in the first list is constrained by a predefined value M;

-populating the second list if the number of symbols n in the first list is smaller than M. The total number of symbols in the second list is constrained by (M-n) such that the total number of symbols in both lists does not exceed M. The coefficients populating the second list are ordered by their position in the raster order, with the magnitude not being greater than T ₁ 。

The context in which a symbol is encoded depends on whether it belongs to the first list or the second list (the difference of step 3 in table 3).

The context in which the symbol prediction error is encoded is determined from the list to which the symbol belongs. The result of this process 802 is a set of candidate locations of coefficients and a context Cx associated with those coefficient locations _j . Thereafter, a special mechanism, such as the mechanism of one of US'444, selects which of the locations belonging to the aggregate symbol prediction to perform (i.e., selects the M coefficients of the symbol prediction). For those unselected locations, a conventional symbol encoding process 809 is performed. In other cases (for the selected M coefficients), symbol prediction 806 is performed according to fig. 7 and equations 2 and 3 described above.

Specifically, in order to predict the symbol 806, reconstructed pixels of the neighboring block are required. The pixels of the prediction signal of the current reconstruction block 804 are subtracted from the neighboring pixels provided by the neighboring pixel acquisition process 803. The 1D transform 805 provides t used in equation 2 _n And v _m . The prediction transform residual sign 806 includes calculating q according to equation 3 _n And o _n The cost function value is obtained according to equation 2.

The calculation in the transform domain is simpler because it is not t _n And v _m Is used to calculate the cost function. Instead, the method uses only a number of coefficients belonging to a particular row and/or a particular column. The row and column are the corresponding indices of the coefficient positions whose sign is being predicted. The cost calculation block 807 generates a set of symbols that minimizes the cost function of equation 2. A sign prediction error is then calculated, which corresponds to the difference between the actual sign of the M coefficients and the predicted sign of the coefficients. The context Cx is utilized in process block 808 _j Entropy encoding the symbol prediction error. The resulting bits are combined with the result of the conventional symbol encoding 809 and embedded in the code stream in process 811.

In other words, in addition to the decoder, there is provided an image block encoding apparatus including a processing circuit for: predicting a sign of a plurality of coefficients of a transformed image block according to a cost function, wherein the cost function comprises a transformed difference between neighboring pixels adjacent to the transformed image block and predictions of the neighboring pixels calculated from a prediction signal of the image block; reconstructing the sign of the plurality of coefficients from the predicted sign.

The plurality of coefficients may correspond to the M coefficients mentioned above. The value of M may be fixed or may be provided by the number of coefficients whose magnitude exceeds a certain threshold. It should also be noted that the plurality of M is less than all of the non-zero coefficients in the above example. In principle, however, symbol prediction can also be applied to all non-zero symbols. According to one advantageous implementation, the M coefficients whose sign is predicted are the M coefficients of maximum amplitude among the non-zero coefficients of the current block.

The above-mentioned coefficient encoding may be performed as follows: determining differences between the signs of the plurality of coefficients and the predicted signs of the plurality of coefficients as sign prediction errors; the determined symbol prediction error is then inserted into a coded stream comprising coded image blocks. The encoded bitstream may be a bitstream that also includes image data of the current block and other signaling information regarding the mode in which the current block, and/or other blocks of a still or video image, are encoded.

As also described with reference to symbol decoding, the cost function of an embodiment includes a sum of squared transform differences between neighboring pixels of the neighboring transformed image block and predictions of the neighboring pixels calculated from the prediction signal of the image block. Specifically, in this embodiment, the processing circuit of the encoding apparatus may be configured to: predicting a sign of a plurality of coefficients of a transformed image block, comprising: calculating a transform difference between neighboring pixels adjacent to the transformed image block and predictions of the neighboring pixels calculated from the prediction signal of the image block; calculating a cost function, wherein the cost function comprises a sum of squared transform differences between the transform difference values and the transformed image blocks, wherein the transformed image blocks are reconstructed from each hypothesis of a symbol in a set of hypotheses; the assumption of a sign that minimizes the cost provided by the cost function is chosen as the predicted sign.

The set of hypotheses consists of a combination K of M symbols, which may include all possible combinations (i.e., powers of 2 of the M combinations). The encoder may further comprise processing circuitry to: the remaining transform coefficient symbols, except for the M symbols, are encoded by binary encoding-1 being positive and 0 being negative, and vice versa.

The cost function F can also be provided in the same way as the decoder, i.e. Wherein t is _n ＝Trans1D(T _n )，q _n ＝Trans1D(Q _n )，v _n ＝Trans1D(V _n )，o _n ＝Trans1D(O _n ) The method comprises the steps of carrying out a first treatment on the surface of the Wherein, trans1D ()' is one-dimensional orthogonal transformation, T _n ＝2X _n,-1 -X _n,-2 -P _n,0 ]，V _m ＝[2Z _-1,m -Z _-2,m -P _0,m ]，Q _n ＝R _n,0 ，O _m ＝R _0,m The method comprises the steps of carrying out a first treatment on the surface of the Wherein P is a prediction signal, X and Y are adjacent pixels, and N and M are the blocks where the prediction symbols are located, respectivelyHeight and width.

The processing circuit of the encoder is further configured to: the symbol prediction error is encoded using Context-adaptive binary arithmetic coding (CABAC) (Context-Adaptive Binary Arithmetic Coding). However, the present invention is not limited to the use of CABAC. Instead, any entropy coding is employed that adapts the length of the codeword to the probability of encoding the corresponding symbol with the corresponding codeword (e.g., CAVLC or any VLC coding).

According to one embodiment, there is provided an image block encoding apparatus including processing circuitry to:

prediction 1320 of the sign of the transform coefficients obtained by transforming 1300 the image block signal using one of a plurality of transforms;

-determining 1340 a symbol prediction error, the symbol prediction error indicating whether the symbol prediction is correct;

-selecting 1360 a context for entropy encoding the symbol prediction error according to one of the plurality of transforms;

-encoding 1380 the symbol prediction error by applying entropy coding using the selected context.

Symbol prediction of transform coefficients may be performed as described above with reference to fig. 6, i.e., according to a cost function associated with smooth transitions and their neighborhood of blocks with predicted symbols. In general, symbol prediction corresponds to the most probable assumption that the symbol bit value minimizes the cost function (in particular minimizing the cost, which may correspond to maximizing the smoothness, thus minimizing the difference between the block with the predicted symbol and the neighborhood). Cost function calculation can be performed in a frequency domain, and the calculation complexity is reduced.

An image block may refer to a current image block of a current image, i.e. a currently processed image block. The processing (encoding) of individual tiles described herein may be applied on a block basis for each tile in an image or some tiles in an image.

It should be noted that in the present description, for example, the image block signal may be a prediction error signal, corresponding to the residual block 105 in fig. 1. In other words, the transform is applied to the residual, not directly to the pixel values. This is also a transform coding scheme commonly applied in several codecs such as h.264/AVC or h.265/HEVC. However, in general, the invention is also applicable to direct conversion of image block 103, i.e. luminance and/or chrominance pixels or any other representation of an image block.

The symbol prediction error indicating whether the symbol prediction is correct may be a difference between the symbol prediction and the true symbol of the transform coefficient. But also binary values, e.g. if the symbol prediction is correct, the value may be zero (corresponding to the zero difference between the prediction and the original symbol); if the symbol prediction is incorrect, the value may be 1 (corresponding to a difference of 1 between the prediction and the original symbol), or other result. If the prediction is good, most of the time the sign prediction error is equal to zero (generally referred to as the prediction correct value). Thus, entropy encoding may be more efficient than fixed length encoding.

Context selection may be performed based on the association between the different transforms and the respective contexts. For example, each of the plurality of transforms is associated with a single context. The association specifies for each transform a context to be selected or used for entropy encoding of the corresponding transform coefficient. It should be noted that the terms "transform" and "transformation" are used herein as synonyms.

Furthermore, the context selection may be performed according to the association between different transformations and respective multiple contexts (sets). For example, each of the plurality of transforms may be associated with a respective plurality of contexts. In addition to the transformations used, multiple contexts are applicable if other parameters or criteria are used to determine the associated context. In particular, the entropy encoded context may then be selected from a corresponding plurality of contexts according to the other parameters or criteria. The association specifies a plurality of contexts (sets) for each transform, from which the context is selected (according to the other parameters) and used for entropy encoding the corresponding transform coefficients.

For example, to implement the association, a look-up table (LUT) may be used, as shown in fig. 11. Specifically, FIG. 11 shows in the first row of Table NFrom T ₀ To T _N N is an integer greater than 1. In the first column of the table, the same transform T for horizontal transforms is shown ₀ To T _N . It should be noted, however, that the number of rows of the table may be different from the number of columns, and in general, the horizontal and vertical transforms may be different. The cells in the table are denoted as CTX _ij (i and j represent the integer indices of the table rows and columns, respectively), the context of residual symbol prediction error coding to be selected for an image block that has been transformed by the transform may be directly represented. For example, the context may be represented by an occurrence probability p (1) of 1 or an occurrence probability p (0) of 0, where p (0) +p (1) =1.

However, in some embodiments, the cells of the table do not directly include the context. In contrast, CTX _ij May be a pointer to a memory storing the applicable context. For example, CTX _ij May be an index (integer) pointing to a context in a table stored in memory. Such a representation is particularly advantageous in case other parameters are used in addition to the transform to select the context for encoding the residual prediction error. Specifically, pointers used as indexes may point to the rows of table 1 and table 2. Table 1 may hold contexts associated with index values of first values of other parameters. Table 2 may hold contexts associated with index values of second values of other parameters. Accordingly, pointer (index) CTX _ij Contexts in the context table are identified, and other parameters identify the context table. It should be noted, however, that the above examples are only a few of the possible implementations. Alternatively, one or more other parameters may define pointers to tables, which may be determined from a plurality of tables corresponding to respective transforms (e.g., a combination of vertical and horizontal transforms).

Transforming the image block to obtain coefficients thereof may be a horizontal transform T _HOR Vertical transform T _VER Is a combination of (a) and (b). The top of the table of fig. 11 is the horizontal transform (left side) and the vertical transform (right side). Specifically, the rows are horizontally transformed and the columns of the image blocks are vertically transformed. The effect of the different combinations of horizontal and vertical transforms is shown in fig. 15. FIG. 15 also shows thatWith some exemplary transforms belonging to the above-described DCT-II, DCT-IV, DST-IV, and identity transforms, each pixel value of the input image is unchanged and mapped directly to a corresponding pixel value of the output image. However, the invention is not limited to these transforms, but may also use discrete fourier transforms (discrete Fourier transformation, DFT) or fast fourier transforms (fast Fourier transformation, FFT) or other transforms in any integer form thereof. The transformations shown in table 1 above may be used in addition to or instead of. In general, any transform is applicable, such as an orthogonal transform or an orthogonal linear transform. Applying a partitionable 2D transform may provide some advantages, i.e. the transform may be divided into a horizontal transform and a vertical transform, which when applied after each other, result in a 2D transform. However, the invention is not limited thereto and in general any 2D transform may be used. In the case of application of non-partitionable 2D transforms, each of these transforms is associated with a particular context set.

Entropy codes assign shorter codewords to more frequent source symbols. Such encoding is efficient if the source symbols are unevenly distributed. In the case of symbol prediction errors, the source symbol may be either 0 or 1, indicating correct and incorrect predictions of the symbol, respectively. The context represents the probabilities of 0 and 1. Since there are only two possible symbols, the context can be provided by a probability of 0 or 1 (since they add up to 1).

The symbol prediction error is entropy encoded/decoded using the selected context. In particular, but not exclusively, a Context-adaptive binary arithmetic coding (CABAC) engine may be used for entropy encoding/decoding using a selected Context. The CABAC context may then be limited by the given combination of vertical and horizontal transforms belonging to the adaptive multi-transform (Adaptive Multiple Transform, AMT)/enhanced multi-transform (Enhanced Multiple Transform, EMT) described above.

After completing the residual symbol prediction step and obtaining a symbol prediction error value indicating whether the symbol prediction of the quantized residual transform coefficients is correct, in one exemplary implementation, the following steps are performed to obtain context to encode the symbol prediction error:

_– Adjacent pixels are scanned. I.e. the available pixels adjacent to the current block of symbols to be estimated are determined and provided for further processing.

_– As in the above embodiments, the forward 1D conversion is performed on the differential signal between the rows and/or columns of the adjacent pixels and the prediction signal. The transform corresponds to a transform used to obtain coefficients.

_– Coefficients are selected in the quantized residual signal for which prediction bits are desired. Namely, select M>1 and less than the number of non-zero coefficients in the block. Alternatively, all non-zero coefficients may be used for symbol prediction.

_– The hypothesis of the bit values (corresponding to the coefficient symbols) is checked and the code prediction error signal corresponding to the most probable hypothesis is encoded.

_– The CABAC context is selected limited to a given combination of vertical and horizontal transforms belonging to AMT/EMT, or is typically selected for the transform used to obtain coefficients and optimize the cost function of symbol prediction.

_– The symbol prediction error is encoded/decoded using a CABAC engine having a selected context.

According to one embodiment, the context is selected 1360 according to an unsigned value of the transform coefficient. In other words, the values of the unsigned transform coefficients determine the other parameters described above. It should be noted that the transform-sensitive contexts described in this embodiment are generally applicable to transform coefficients, whether (and how) they are quantized after transformation, and whether they are obtained by directly transforming image pixels/pixel values or residuals. However, examples involving HEVC and its improvements may generally apply transforms to residuals.

For example, the unsigned value may be a value based on a norm of the transform coefficient or a value based on an absolute value of the transform coefficient.

According to one embodiment, the processing circuit is further configured to: predicting each of 1320M transform coefficients, M being an integer, greater than 1 and less than a number of transform coefficients obtained by transforming the image block; determining 1340 a symbol prediction error for each of the M transform coefficients; dividing the M transform coefficients into two lists through a threshold value, wherein a first list comprises transform coefficients with absolute values larger than the threshold value, and a second list comprises transform coefficients with absolute values equal to or smaller than the threshold value; selecting 1360 a context for entropy encoding a symbol prediction error of the transform coefficients of the M transform coefficients according to whether the transform coefficients of the M transform coefficients belong to the first list or the second list; the symbol prediction errors of the transform coefficients of the M transform coefficients are encoded 1380 by applying entropy encoding using the selected context. In other words, if the coefficient belongs to list 1, the prediction error of its symbol is encoded using context 1. On the other hand, if the coefficient belongs to list 2, the prediction error of its symbol is encoded using context 2. As described above, the transform has pre-selected context 1 and context 2. For example, pointers for contexts 1 and 2 are obtained from the transform.

The threshold is a predetermined or predefined threshold, which may be included in the criteria. The threshold may be set empirically. Its purpose is to distinguish between "natural" discontinuities (caused by the natural nature of the picture) and "artificial discontinuities" (in this case caused by the wrong sign of the transform coefficient). For example, the threshold may be obtained by trying different values and observing their effect on the coding efficiency on a given image database. The threshold value is not necessarily defined as a constant value. May be transmitted in a picture-level code stream, for example, in a sequence parameter set (sequence parameter set, SPS) or a picture parameter set (picture parameter set, PPS).

The list may be an RSP list, as follows:

_– the HP list (high probability list) is a list comprising positions of such quantized transform coefficients having an amplitude above a threshold,

thus, the signs of these quantized transform coefficients can be correctly predicted with a large probability;

_– the LP list (small probability list) is a list comprising the positions of such quantized transform coefficients with magnitudes below a threshold value, and therefore the signs of these quantized transform coefficients can be correctly predicted with a small probability.

However, it should be noted that there may be more than two lists.

For example, returning to FIG. 11, vertical T is used _VER And level T _HOR Transformed index, a pointer CTX can be obtained from a 2D LUT _KL (K and L are indices of table rows and columns). The pointer CTX _KL Two (possibly different) context CTX for entropy encoding/decoding of symbol prediction errors of residual transform coefficients belonging to the respective HP and LP lists may be accessed _KLH And CTX _KLL . In general, by CTX _KL The number of available contexts is equal to the number of RSP lists. Pointer CTX _KL May be used to select a context corresponding to the RSP list to which the residual transform coefficient belongs.

In this embodiment, the symbols of some of the transform coefficients obtained by transforming the image block are encoded using CABAC bypass mode. That is, the number of non-zero coefficients is at least 2 less than the total number of coefficients obtained by transforming the image block, and entropy decoding is not performed (see also the description above in connection with fig. 8). Therefore, the number of transform coefficients remaining to be entropy-encoded is M. Thus, M is at least 2 and is less than the total number of coefficients obtained by transforming the image block.

In general, there may be more than two lists. The transform coefficients may then be distributed in a list according to the coefficient values. Thus, in this case, there may be several thresholds defining corresponding unsigned coefficient value ranges. For example, transform coefficients greater than a first threshold may be included in a first list, transform coefficients less than a second threshold may be included in a second list, and the remaining transform coefficients may be included in a third list. According to this example, in addition to one of a plurality of transforms for acquiring corresponding transform coefficients from an image block, a context for entropy encoding the transform coefficients may be selected according to whether the transform coefficients of the M transform coefficients belong to a first list or a second list or a third list. In the case of more than two lists, each transformation is associated with one context of each list, and each of these contexts is in turn associated with one of the lists. Thus, each combination of transforms and lists is associated with one of at least two different contexts. In this case, the same context may have multiple transforms and list combinations associated with it. In particular, multiple combinations of the same transformation with different lists may be associated with the same context. Furthermore, multiple combinations of different transforms and the same list may be associated with the same context.

In summary, for each of the M transform coefficients, if the corresponding transform coefficient is in the first list, the first context may be used; the second context may be used if the corresponding transform coefficient is in the second list. Thus, the first context and the second context are pre-selected by a transform for obtaining the respective transform coefficients from the image block. In other words, each transformation is associated with both contexts, each of which is in turn associated with one of the two lists. This is also illustrated below the LUT table of fig. 11. The list may be ordered by unsigned coefficient value (in ascending or descending order). It should be noted that the term "list" herein generally refers to an ordered (sorted) data set.

According to one embodiment, a context is selected 1360 depending on whether the prediction type of the image block is intra prediction or inter prediction.

The selection criteria of the present embodiment may be added to or alternatively selected on the basis of the above-described HP/LP list or other selection criteria. Thus, CTX pointers may point to different context tables for intra-prediction and inter-prediction, or to different tables for a combination of prediction types and correlations with list 1 or list 2. For example, in the case of two lists, the HP and LP lists, the CTX pointer may point to four different tables for each transformation, that is, each transformation may be subsequently associated with a different context table.

The present embodiment suggests selecting a different set of contexts depending on whether a block is predicted using temporal (inter) prediction (i.e., using one or more different frames) or intra prediction (using neighboring pixels of the same image as the predicted block). There are two reasons for making this distinction:

(i) Inter-prediction blocks have fewer quantized transform coefficients than intra-prediction blocks, because inter-prediction generally yields better predictions.

(ii) Intra-prediction blocks use neighboring pixels for predicting the block and estimating discontinuities. Thus, the inter prediction block predicts a symbol using additional input data compared to intra prediction.

More specifically, the input data for reconstructing the inter-block pixels includes pixel points of the reference frame and the residual signal. On the other hand, the input data for reconstructing the intra-block pixels includes neighboring pixel points and residual signals.

Discontinuities between neighboring pixels and reconstructed pixels are estimated that are related to symbol hypotheses. In fact, for intra-predicted blocks, the reconstructed pixels are predicted from neighboring blocks, i.e., those used to estimate the discontinuity. This may lead to an estimate bias because part of the signal is generated from the compared signals.

In the case of inter prediction, neighboring pixels are not used to acquire a prediction signal, but are used to estimate discontinuities. In this sense, inter-predicted blocks have more information (and less deviation) than intra-predicted blocks.

According to one embodiment, if the prediction type of the image block is intra prediction, a context is selected 1360 according to an intra prediction mode used to predict the image block.

In this embodiment, the context used for entropy encoding the symbol prediction error is selected according to the intra prediction mode used, i.e., a different context may be selected for each different intra prediction mode specified by the video codec. In particular, different contexts may be provided for each particular direction of intra-prediction, DC intra-prediction, planar (plane/planar) prediction modes, and different contexts may be provided for each non-directional mode. In other words, the intra prediction mode used to predict the image block may be DC prediction, planar prediction, or a direction prediction mode having specific directions, where each specific direction may be considered a different intra prediction mode (implying a different context that may be selected).

The selection criteria of the present embodiment may be added to or alternatively selected on the basis of the above-described HP/LP list or other selection criteria. Thus, the CTX pointer may point to a different context table for each intra prediction mode. These tables may look similar to table 2 but are populated with contexts instead of transform sets.

It should be noted that alternatively, a combination of parameters such as the LP/HP list, inter prediction/intra prediction and/or prediction modes (and/or possibly other parameters) listed herein may be used as a combination associated with pointers to tables with contexts, one for each transformation. There are only two possible implementations, but in general, the association between parameters/parameter combinations and the respective contexts may be implemented in any way.

By specifying a context for each intra-prediction mode, it can be considered whether the symbol prediction probability (sign prediction probability, SPP) depends on the intra-prediction mode.

According to one embodiment, the sign of the transform coefficients is predicted 1320 according to a cost function, the cost function including estimating discontinuities along the boundaries of the image block; the context is selected 1360 according to the number of block edges or the number of neighboring pixels that can be used to estimate the discontinuity on the boundary of the image block.

Thus, by evaluating the cost function corresponding to maximizing smoothness, discontinuities are minimized.

In comparison to the previous example, it is worth noting that there are other segmentation frameworks where there may be more than two edges for intra/inter prediction and further symbol prediction. For example, bottom and/or right side blocks may be available in addition to or with the top and left side blocks of the above examples. In practice, the availability of blocks depends on the order of processing of the blocks in the image. In general, the block coding (processing) order can be adaptively determined at the encoding end and transmitted to the decoder. Thus, adjacent pixels of a coded block may be located on more than two sides of the block. In addition, some edges may be used in part for prediction. Thus, in another possible embodiment, the context is typically selected according to the number of edges of the block or the number of neighboring pixels that can be used to estimate the discontinuity along the boundary of the current block. In particular, if hierarchical segmentation is applied, a block may have one or more neighbors on one of its sides. For example, an image block a of size 16×16 may have two blocks of size 8×8 on its left boundary. On its upper boundary, block a may have three neighbors: one 8 x 8 block and two 4 x 4 blocks. Some of which may be used for prediction and/or cost function calculation, but some are not necessarily available.

According to one embodiment, the 1360 context is selected according to the ratio of edge pixels available for estimating discontinuities to the total number of pixels belonging to the image block.

The segmentation information also affects the context selection process. For example, if the video codec uses quadtree partitioning, the resulting encoded block is square in shape. In this case, two factors affecting the probability are the size of the block and the availability of neighboring pixels. To take these two factors into account, the present embodiment suggests using a ratio of edge pixels to the total number of pixels belonging to a block (edge area ratio). Table 4 provides a calculation example of this ratio. Specifically, if the current image block, i.e., the block whose sign is to be predicted, is 16×16 in size and is only on one side (e.g., left or top), the number of pixels having neighboring pixels available for prediction is 16, and the number of pixels in the 16×16 block is 256. The corresponding ratio was 16/256, resulting in about 0.06. On the other hand, if both sides of the block have neighboring pixels available for prediction, the number of pixels having neighboring pixels available for prediction is 2×16-1, which is equal to 31. The resulting ratio was 31/256, resulting in about 0.12. Similarly, table 4 may be expanded by rows assuming 3 or 4 sides have adjacent pixels available.

Table 4: edge area ratio

Context selection may be performed by mapping the ratio to a context index. An exemplary mapping is shown in table 5. Table 5 shows the correlation between the context index and the corresponding edge area ratio range. Here, parentheses indicate that the range limitation (e.g., 0, 0.1, 0.2, 0.3) is not included in the range, and brackets indicate that the range limitation (e.g., 0.1, 0.2, 0.3, 0.5) is included in the range.

Table 5: selecting a context according to an edge area ratio

Context index	0	1	2	3
					Edge area ratio	(0；0.1]	(0.1；0.2]	(0.2；0.3]	(0.3；0.5]

The present embodiment is applicable not only to square blocks but also to rectangular shapes.

For rectangular blocks, table 4 may additionally use aspect ratio, i.e., width of the block divided by height of the block. In this case, the edge area ratio will be selected based on one of the length (e.g., block width) and aspect ratio of the block edge. Several mappings of edge area ratios to context indices may be defined (as shown in table 5), and corresponding mappings may be selected for aspect ratio values.

According to one embodiment, the 1360 context is selected based on whether the tile is located at the left or upper boundary of the tile where the tile is located.

Typically, a slice is a portion of an image that can be decoded independently of other slices in the same image. This is achieved by (1) restarting entropy coding at the beginning of a slice and (2) using intra prediction without crossing slice boundaries. Therefore, to achieve the design goal of providing independent slices, not performing RSP across slice boundaries may help optimize costs.

Specifically, symbol prediction is performed according to a calculation cost estimation function. This function estimates discontinuities along the boundary of the current block. Obviously, the cost estimation function may be calculated for only the boundaries available for neighboring pixels. The availability of neighboring pixels is determined by the block coding order. For example, in the case of a z-order block scan, the current block may have only left and upper available neighboring pixels. The left or top adjacent pixels of the block to the left or top of the alignment patch are not available. Therefore, the cost estimation function uses a smaller number of pixels for these blocks and therefore the probability of correct symbol prediction for these blocks is lower.

In other words, even though image blocks may have adjacent pixels, they are not used if they belong to a different slice than the slice in which the image block is located. Thus, context selection may consider the position of an image block at a slice boundary as an indication of prediction quality, for use in context selection.

In an embodiment of the context selection for symbol prediction errors, the use of such context selection takes into account the dependency of the position. It is suggested to select different contexts for the blocks aligned with the left and top sides of the tile and the rest of the tile.

According to one embodiment, in particular, an image block is predicted using a DC intra prediction mode or a planar mode; the picture selection 1360 context is predicted according to whether DC intra prediction mode or planar mode is used.

If the blocks have natural discontinuities (e.g., object edges or texture regions), their pixel values are less correlated to neighboring pixels, so these blocks are more likely to be predicted using the non-directional modes (i.e., DC modes or planar modes) used in this embodiment. Two specialized contexts may be provided: the DC intra prediction mode selects one context when predicting a block and another context when using the planar mode. Alternatively or additionally, each other mode may provide other context if the video codec specifies more non-directional modes than DC-only, planar modes.

Alternatively, the planar and DC intra prediction modes may use the same context (context 1 or context set 1) and may be different from the context used for directional interpolation (context 2 or context set 2).

According to one embodiment, the sign of the transform coefficients is predicted 1320 according to a cost function, wherein the cost function comprises transform differences between neighboring pixels of a neighboring transformed image block and predictions of the neighboring pixels calculated from a prediction signal of the image block. Transform domain optimization has been described in detail in fig. 4-8 above.

According to one embodiment, each of the plurality of transforms is associated with one of a plurality of contexts, the context being selected 1360 from the plurality of contexts such that the one of the plurality of transforms is associated with the selected context.

If additional parameters or criteria are used to select a context, each of the plurality of transforms is associated with one of a plurality of context sets. Specifically, in this case, each of the plurality of transforms is associated with a context set. A context is then selected from the set of associated contexts in accordance with the further parameter or criteria.

For example, as shown in FIG. 11, if other parameters are used to select a context, each of a plurality of transforms may be associated with a pointer.

According to one embodiment, a context for entropy encoding the symbol prediction error is selected 1360 according to one of the plurality of transforms each associated with one of a plurality of contexts, wherein a first transform of the plurality of transforms is associated with a same context of a second transform of the plurality of transforms if the probabilities of predicted symbol values of the first transform and the second transform differ by less than a predetermined value.

The predetermined value is a design parameter that is affected by the total number and characteristics of the transforms used with the codec and the properties of the entropy encoder. In particular, the "window length", i.e. how many previously encoded values are taken into account by the probability model of the entropy encoder. Thus, the predetermined values may be various predetermined empirical values tested using a database of training video sequences.

In further embodiments, several contexts are grouped together. In particular, if intra prediction modes close to horizontal and vertical intra prediction modes are used to predict a block, a connection context (DC and planar mode grouping, several similar or all directional mode groupings, etc.) may be used.

Further, the context selected from the cells shown in fig. 12 is not a completely unique context. If the probabilities of the correct symbol predictions are similar, several contexts can be combined into one context. In FIG. 12, the two CTX units are combined to form a single CTX unit ₀₀ Cells (0, 0) and (1, 1), which may be contexts or the same pointer to the same context, are shown.

Thus, for example, a separate pair of contexts is not required for each combination of vertical and horizontal transforms. If the probability of the speculative (predictive symbol) value (e.g., P _kl And P _mn ) Sufficiently close (less than ε) for these combinations, the corresponding contexts can be merged:

abs(P _kl –P _mn )<ε

here, the indices k, l, m and n mark the corresponding horizontal and vertical transforms. Thus, P _kl Representing the probability of the kth horizontal transform being combined with the ith vertical transform.

However, the above formula is not limited to a combination of different horizontal transforms and vertical transforms. For some sufficiently small values epsilon,

abs(P _i –P _j )<ε

any two contexts with corresponding probabilities may be combined, where i and j are indices in different contexts, including contexts and/or other parameters for transformation.

The threshold ε in the above formula is determined experimentally because its choice depends on various factors including, but not limited to: the total number of contexts used by the video codec, the working rate point (high, medium, low), the efficiency or other properties of the entropy encoder (specifically "window length", i.e. how much previously encoded values the probability model of the entropy encoder takes into account), the maximum number of symbols per block prediction, etc., as described above.

According to one embodiment, the prediction 1320 predicts a sign of a transform coefficient obtained by transforming 1300 an image block signal using one of a plurality of transforms, wherein the transforms include a horizontal transform and a vertical transform; the selection 1320 of the context is based on a combination of the horizontal transform and the vertical transform of the transform.

According to one embodiment, each of the plurality of transforms is a combination of a horizontal transform and a vertical transform, the processing circuit further being configured to: the context is selected 1360 such that the combination of the horizontal transform and the vertical transform is associated with the selected context.

The choice of the context of the symbol prediction error coding is determined by the symbol prediction probability (sign prediction probability, SPP), i.e. the frequency of correctly predicting the symbol. In fact, the SPP differences of different horizontal and vertical transform combinations result in other (CABAC) contexts that have to be introduced.

This example is shown in fig. 11. As shown, each combination of horizontal and vertical transforms may be associated with a pointer as described above.

However, in general, the number of horizontal transforms and vertical transforms may be different, and the manner of transformation of the horizontal transforms and vertical transforms may also be different. For clarity of illustration, the first row may be from T ₀ To T _N Run whileThe first column may be from T' ₀ To T' _M And (5) running.

According to one embodiment, the number of mutually different contexts of said combination of vertical transforms and said horizontal transforms is smaller than the number of said combinations of vertical transforms and said horizontal transforms, a first combination of vertical transforms and horizontal transforms being associated with the same context of a second combination of vertical transforms and horizontal transforms.

Newly introduced context number N _CTX Calculated by the following inequality:

N _CTX ≤N _{RSP_LISTS} *N _{TR_COMB}

wherein N is _{RSP_LISTS} Is the number of RSP lists (in this particular case, N _{RSP_LISTS} =2, i.e. different contexts of coefficients in HP and LP lists), N _{TR_COMB} Is the number of possible combinations of horizontal and vertical transforms that can be used (e.g., N _{TR_COMB} =5), "x" denotes multiplication.

Shown in fig. 12, wherein (T ₀ ，T ₀ ) Table entry and (T) ₁ ，T ₁ ) The table entries are the same and the table entries,

CTX ₀₀ ＝CTX ₁₁ 。

however, in contrast to fig. 12, the number of vertical transforms and horizontal transforms cannot be the same. Furthermore, the horizontal transformation and the vertical transformation may be different (see above).

In general, i.e. if other parameters or criteria are used to determine the associated context or to select the context for entropy encoding, the above formula can be extended by other factors on the right. The factors correspond to the number of values of other parameters used to determine the context.

According to one embodiment, context-adaptive binary arithmetic coding (CABAC) is used for entropy coding 1380. However, any form of context-adaptive entropy coding may be used, such as context-adaptive variable length coding (Context Adaptive Variable Length Coding, CAVLC) or different types of entropy coding (VLC).

According to one embodiment, there is provided an apparatus for decoding an image block from a code stream, the apparatus comprising processing circuitry to:

predicting 1420 a symbol of transform coefficients to be inverse transformed using one of a plurality of transforms to obtain an image block signal;

-selecting 1440 a context for entropy decoding the symbol prediction error according to one of the plurality of transforms, the symbol prediction error indicating whether the symbol prediction is correct;

_– decoding 1460 the symbol prediction error by applying entropy decoding using the selected context;

_– the sign of the transform coefficient is determined 1480 based on a prediction sign of the transform coefficient and a decoded sign prediction error of the transform coefficient.

A context for entropy decoding the symbol prediction error is selected according to the corresponding inverse transform.

The corresponding inverse transform may be a transform in which the decoder obtains an image block or image block signal from the transform coefficients. The corresponding inverse transform may also be an inverse transform of the transform used by the encoder to obtain transform coefficients from the image block or image block signal. The context for entropy encoding the symbol prediction error may then be selected according to the transform used to obtain the transform coefficients from the image block.

The inverse transform may be one of a plurality of inverse transforms. In general, multiple inverse transforms may be equivalent to multiple transforms and vice versa. In particular, the transform may be an inverse transform of another transform, as in the case of a different DCT.

Each of the plurality of transforms may be associated with one of a plurality of contexts. The context selected for entropy decoding is the corresponding context associated with the relevant transform, which is the transform related to the coefficients whose symbols are to be entropy decoded.

In addition, other parameters or criteria may be used to determine (i.e., select) the context in which to entropy decode. In this case, each of the plurality of transforms may be associated with a plurality of contexts. A particular context for entropy decoding is selected from a plurality of contexts associated with the transform according to the other parameters or criteria.

The contexts associated with two different transforms may be different or the same (e.g., merged). In the case of multiple contexts, the multiple contexts associated with two different transforms may be different, identical (e.g., merged), or have overlapping contexts (i.e., only some of the same contexts—not necessarily for the same other parameters or criteria).

The sign of the transform coefficient is determined based on a prediction sign of the transform coefficient and a decoded sign prediction error of the transform coefficient. That is, the decoded symbol prediction error indicates whether the symbol prediction is correct. In combination with the predicted symbols, the correct sign of the respective corresponding coefficient can be inferred.

The apparatus provided by this embodiment may further include a circuit for inverse transforming the transform coefficients and their signs parsed from the code stream, the signs of the transform coefficients being reconstructed from the prediction error and the prediction as described above.

Fig. 13 and 14 illustrate steps of a method for encoding and decoding a prediction error of a residual symbol prediction. It should be noted that the method according to the invention comprises all the steps described above in connection with the processing circuit.

In step 1300 of fig. 13, an image block is first transformed, and transform coefficients obtained by the transformation are quantized. The image block encoding method includes the steps of:

_– the prediction 1320 transforms 1300 the sign of the transform coefficient obtained by transforming the image block signal using one of a plurality of transforms;

_– determining 1340 a symbol prediction error, the symbol prediction error indicating whether the symbol prediction is correct;

_– Selecting 1360 a context for entropy encoding the symbol prediction error according to one of the plurality of transforms;

_– by using a selected context applicationEntropy encoding encodes 1380 the symbol prediction error.

Similarly, fig. 14 shows a method of decoding an image block from a code stream, comprising the steps of:

_– predicting 1420 a symbol of transform coefficients to be inverse transformed using one of a plurality of transforms to obtain an image block signal;

_– selecting 1440 a context for entropy decoding the symbol prediction error according to one of the plurality of transforms, the symbol prediction error indicating whether the symbol prediction is correct;

In step 1400, the code stream is parsed. Such parsing operations may include parsing encoded unsigned coefficient values from the code stream. For example, after parsing the coefficient values, a list of LP/HP may be constructed and the corresponding symbols encoded as described above.

Fig. 9 shows an exemplary implementation of the embodiment described above in the encoder in fig. 1. Specifically, the newly introduced module 122 performs symbol prediction and the modified module 970 performs entropy encoding.

The location of the symbol encoded using the CABAC context is defined based on the magnitude of the quantized coefficients, and thus the symbol prediction 122 uses the result of the quantization process 108. The entropy encoding module 970 may include the following two modifications:

_– the symbol prediction error 123 is encoded instead of the symbol values of the set of M given positions within the block of quantized transform coefficients 109.

_– New contexts are introduced and applied to encode the symbol prediction error 123 in the entropy encoder 970 to produce an output 971.

_– Furthermore, according to one embodiment, the context of entropy encoding is selected according to the transformation applied in block 106 and possibly other criteria as described above.

Fig. 10 shows an exemplary implementation of the embodiment described above in the decoder in fig. 2.

The decoder in fig. 10 has a corresponding new symbol prediction module 122 and a modified entropy decoding module 1004. For example, in some embodiments, entropy decoding may be performed based on a context selected according to the transform applied in block 106 and possibly other criteria as described above.

In particular, the entropy decoding module 1004 parses from the stream coefficients and other syntax elements that may be used for symbol prediction. Accordingly, the decoding module 1004 may perform step 1400. Based on the parsed information, the coefficient symbols are predicted in block 122, and block 122 may perform the step of symbol prediction 1420. Based on the prediction (not shown in fig. 10), entropy decoding module 1004 may also select context 1440, decode the symbol prediction error using context 1460, and reconstruct symbol value 1480.

The symbol prediction module 122 is substantially identical to the symbol prediction module used at the encoding end at least in terms of symbol prediction computation (refer to fig. 9), but uses the prediction error signal 123 parsed from the code stream 971. The symbol prediction 123 recovers some (M) symbols of the quantized transform coefficients 209, which quantized transform coefficients 209 are located at (M) positions defined by the intra-block amplitude distribution of the quantized transform coefficients 209. In the parsing process (entropy decoding 1004), only the symbol prediction error signal 123 is recovered, and the symbol itself is not recovered. The embodiment of the present invention does not introduce parsing dependencies because the actual symbol values and positions may be recovered from the prediction error signal 123 in parallel with the dequantization process 210, in particular, after the parsing process 1004 is completed.

The parsed symbol prediction error 123 is added to the predicted symbol in block 122, resulting in a decoded (reconstructed) symbol 209. The prediction symbols are obtained in the same way as in the encoder, using only decoded neighbor blocks, and testing K hypotheses for M symbols.

It should be noted that the present description provides an explanation of an image (frame), but in the case of an interlaced image signal, fields replace images.

Although embodiments of the present invention have been described primarily in terms of video encoding, it should be noted that embodiments of encoder 100 and decoder 200 (and correspondingly, system 300) may also be used for still image processing or encoding, i.e., processing or encoding of a single image independent of any previous or subsequent image in video encoding. In general, only the inter-frame estimation 142 and the inter-frame predictions 144, 242 are not available in the case where the image processing encoding is limited to a single image 101. Most, if not all, other functions (also referred to as tools or techniques) of the video encoder 100 and video decoder 200 are equally applicable to still images, such as segmentation, transformation (scaling) 106, quantization 108, inverse quantization 110, inverse transformation 112, intra-frame estimation 142, intra-frame prediction 154, 254, and/or loop filtering 120, 220, entropy encoding 170, and entropy decoding 204.

Those skilled in the art will appreciate that the "blocks" ("units") of the various figures (methods and apparatus) represent or describe the functions of the embodiments of the invention (rather than necessarily individual "units" in hardware or software), and thus also describe the functions or features of the apparatus embodiments as well as the method embodiments (units = steps).

The term "unit" is used merely to illustrate the functionality of the encoder/decoder and is not intended to limit the present invention.

In providing several embodiments in the present application, it should be understood that the disclosed systems, apparatuses, and methods may be implemented in other ways. For example, the described apparatus embodiments are merely exemplary. For example, the cell division is merely a logical function division and may be other divisions in a practical implementation. For example, multiple units or components may be combined or integrated into another system, or some features may be omitted or not performed. In addition, the mutual coupling or direct coupling or communication connection shown or discussed may be implemented through some interfaces. The direct coupling or communication connection between devices or units may be accomplished electronically, mechanically, or in other forms.

Elements described as separate parts may or may not be physically separated, and parts described as elements may or may not be physical elements, may be located in one location, or may be distributed over a plurality of network elements. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

In addition, functional units in embodiments of the invention may be integrated into one processing unit, or each unit may physically reside in a single unit, or two or more units may be integrated into one unit.

Embodiments of the invention may also include apparatuses, e.g., encoders and/or decoders, including processing circuits, for performing any of the methods and/or processes described herein.

Embodiments of encoder 100 and/or decoder 200, as well as other embodiments, may be implemented as hardware, firmware, software, or any combination thereof. For example, the functions of encoder/encoder or decoder/decoder may be performed by processing circuitry, with or without firmware or software, such as a processor, microcontroller, digital signal processor (digital signal processor, DSP), field programmable gate array (field programmable gate array, FPGA), application-specific integrated circuit (ASIC), etc.

The functions of the encoder 100 (and corresponding encoding method 100) and/or the decoder 200 (and corresponding decoding method 200) may be implemented by program instructions stored on a computer readable medium. The program instructions, when executed, cause a processing circuit, computer, processor, etc. to perform the steps of the encoding and/or decoding method. The computer readable medium may be any medium that stores a program including a non-transitory storage medium, such as a blu-ray disc, DVD, CD, USB (flash) drive, hard disk, server memory available over a network, etc.

Embodiments of the present invention include or are a computer program comprising program code to perform any of the methods described herein when the computer program is executed on a computer.

Embodiments of the invention include, or are a computer-readable non-transitory medium comprising program code that, when executed by a processor, causes a computer system to perform any of the methods described herein.

According to one embodiment, there is provided an image block decoding apparatus including processing circuitry to: predicting a sign of a plurality of coefficients of a transformed image block according to a cost function, wherein the cost function comprises a transformed difference between neighboring pixels adjacent to the transformed image block and predictions of the neighboring pixels calculated from a prediction signal of the image block; reconstructing the sign of the plurality of coefficients from the predicted sign.

In an exemplary implementation, the processing circuit is further to: analyzing a symbol prediction error from the coded code stream; reconstructing the symbol includes adding the parsed symbol prediction error to the predicted symbol.

For example, the cost function includes a sum of squared transform differences between neighboring pixels of the neighboring transformed image block and predictions of the neighboring pixels calculated from the prediction signal of the image block.

Furthermore, in one implementation, the processing circuit is further configured to predict a sign of a plurality of coefficients of the transformed image block, including: calculating a transform difference between neighboring pixels adjacent to the transformed image block and predictions of the neighboring pixels calculated from the prediction signal of the image block; calculating a cost function, wherein the cost function comprises a sum of squared transform differences between the transform difference values and the transformed image blocks, wherein the transformed image blocks are reconstructed from each hypothesis of a symbol in a set of hypotheses; the assumption of a sign that minimizes the cost provided by the cost function is chosen as the predicted sign.

For example, the transformed image block is reconstructed by adding the hypotheses of the symbol to coefficients parsed from the encoded code stream.

Specifically, the cost function F is provided by:wherein t is _n ＝Trans1D(T _n )，q _n ＝Trans1D(Q _n )，v _n ＝Trans1D(V _n )，o _n ＝Trans1D(O _n ) The method comprises the steps of carrying out a first treatment on the surface of the Wherein, trans1D ()'s is one-dimensional orthogonal transformation，T _n ＝[2X _n,-1 -X _n,-2 -P _n,0 ]，V _m ＝[2Z _-1,m -Z _-2,m -P _0,m ]，Q _n ＝R _n,0 ，O _m ＝R _0,m The method comprises the steps of carrying out a first treatment on the surface of the Wherein P is a prediction signal, X and Y are adjacent pixels, and N and M are the height and width of the block where the prediction symbol is located, respectively.

In an exemplary implementation, adjacent pixels may be located at horizontal and vertical boundaries of an image block.

In some embodiments, the processing circuit is further to: the symbol prediction error is decoded using Context-adaptive binary arithmetic coding (CABAC) (Context-Adaptive Binary Arithmetic Coding).

According to one embodiment, there is provided an image block encoding apparatus including processing circuitry to: predicting a sign of a plurality of coefficients of a transformed image block according to a cost function, wherein the cost function comprises a transformed difference between neighboring pixels adjacent to the transformed image block and predictions of the neighboring pixels calculated from a prediction signal of the image block; the symbols of the plurality of coefficients are encoded according to the prediction symbols.

According to an embodiment, there is provided the image block encoding device according to claim 9, wherein the processing circuit is further configured to encode the symbols of the plurality of coefficients, including: determining a sign prediction error as a difference between the sign of the plurality of coefficients and the predicted sign of the plurality of coefficients; the determined symbol prediction error is inserted into a coded stream comprising the coded image block.

In an exemplary implementation, an image block decoding apparatus according to any of claims 9 to 11 is provided, wherein the processing circuit is further configured to: predicting a sign of a plurality of coefficients of a transformed image block, comprising: calculating a transform difference between neighboring pixels adjacent to the transformed image block and predictions of the neighboring pixels calculated from the prediction signal of the image block; calculating a cost function, wherein the cost function comprises a sum of squared transform differences between the transform difference values and the transformed image blocks, wherein the transformed image blocks are reconstructed from each hypothesis of a symbol in a set of hypotheses; the assumption of a sign that minimizes the cost provided by the cost function is chosen as the predicted sign.

For example, the cost function F is provided by:wherein t is _n ＝Trans1D(T _n )，q _n ＝Trans1D(Q _n )，v _n ＝Trans1D(V _n )，o _n ＝Trans1D(O _n ) The method comprises the steps of carrying out a first treatment on the surface of the Wherein, trans1D ()' is one-dimensional orthogonal transformation, T _n ＝[2X _n,-1 -X _n,-2 -P _n,0 ]，V _m ＝[2Z _-1,m -Z _-2,m -P _0,m ]，Q _n ＝R _n,0 ，O _m ＝R _0,m The method comprises the steps of carrying out a first treatment on the surface of the Wherein P is a prediction signal, X and Y are adjacent pixels, and N and M are the height and width of the block where the prediction symbol is located, respectively.

Further, the adjacent pixels may be located at horizontal and vertical boundaries of the image block.

The processing circuit of the decoding apparatus is further configured to: the symbol prediction error is encoded using Context-adaptive binary arithmetic coding (CABAC) (Context-Adaptive Binary Arithmetic Coding).

According to one embodiment, there is provided an image block decoding method including the steps of: predicting a sign of a plurality of coefficients of a transformed image block according to a cost function, wherein the cost function comprises a transformed difference between neighboring pixels adjacent to the transformed image block and predictions of the neighboring pixels calculated from a prediction signal of the image block; reconstructing the sign of the plurality of coefficients from the predicted sign.

According to one embodiment, there is provided an image block encoding method including the steps of: predicting a sign of a plurality of coefficients of a transformed image block according to a cost function, wherein the cost function comprises a transformed difference between neighboring pixels adjacent to the transformed image block and predictions of the neighboring pixels calculated from a prediction signal of the image block; reconstructing the sign of the plurality of coefficients from the predicted sign.

It should be noted that the above embodiments, implementations and examples described in connection with processing circuits in an encoding or decoding device also apply to the above encoding and decoding methods, which may correspond to the steps performed by the respective processing circuits.

In summary, the present invention provides embodiments of symbol encoding and decoding of transform coefficients suitable for image and/or video encoding and decoding, and the like. Specifically, a plurality of the symbols are predicted, and only a prediction error signal is embedded in a code stream. The prediction error signal may have a distribution that may be efficiently encoded with CABAC or other variable length (entropy) encoding. In order to efficiently perform sign prediction, the signs of a plurality of coefficients of a transformed image block are predicted according to a cost function, wherein the cost function comprises transform differences between neighboring pixels of a neighboring transformed image block and predictions of the neighboring pixels calculated from a prediction signal of the image block. Furthermore, embodiments of symbol encoding and decoding of transform coefficients suitable for image and/or video encoding and decoding and the like are provided. Specifically, a plurality of the symbols are predicted, and only a prediction error signal is embedded in a code stream. The prediction error signal may have a distribution that may be efficiently encoded with CABAC or other variable length (entropy) encoding. Further, when an adaptive multi-core transform is used, a context in which the transform coefficient symbols are encoded by an entropy code is selected according to the transform used to obtain the transform coefficients.

REFERENCE SIGNS LIST

FIG. 1

100. Encoder with a plurality of sensors

103. Image block

102. Input (e.g., input port, input interface)

104. Residual calculation [ Unit or step ]

105. Residual block

106. Transformation (e.g. additionally including scaling) [ units or steps ]

107. Transform coefficients

108. Quantification [ Unit or step ]

109. Quantization coefficient

110. Inverse quantization [ Unit or step ]

111. Dequantizing coefficients

112. Inverse transformation (e.g. additionally including scaling) [ units or steps ]

113. Inverse transform block

114. Reconstruction [ Unit or step ]

115. Reconstruction block

116 (column) buffers [ Unit or step ]

117. Reference pixel point

120. Loop filter [ element or step ]

121. Filtering block

130. Decoding image buffer (decoded picture buffer, DPB) [ Unit or step ]

131. Decoding an image

142. Inter-frame estimation (inter-estimation or inter picture estimation) [ Unit or step ]

143. Inter estimation parameters (e.g., reference picture/reference picture index, motion vector/offset)

144. Inter prediction (inter prediction or inter picture prediction) [ Unit or step ]

145. Inter prediction block

152. Intra estimation (intra estimation or intra picture estimation) [ unit or step ]

153. Intra prediction parameters (e.g., intra prediction modes)

154. Intra prediction (intra prediction or intra frame/picture prediction) [ Unit or step ]

155. Intra prediction block

162. Mode selection [ Unit or step ]

165. Prediction block (inter prediction block 145 or intra prediction block 155)

170. Entropy coding [ Unit or step ]

171. Encoding image data (e.g., a code stream)

172. Output (output port, output interface)

FIG. 2

200. Decoder

171. Encoding image data (e.g., a code stream)

202. Input (Port/interface)

204. Entropy decoding

209. Quantization coefficient

210. Inverse quantization

211. Dequantizing coefficients

212. Inverse transform (zoom)

213. Inverse transform block

214. Reconstruction (Unit)

215. Reconstruction block

216 (column) buffer

217. Reference pixel point

220. Loop filter (within loop filter)

221. Filtering block

230. Decoding image buffer (decoded picture buffer, DPB)

231. Decoding an image

232. Output (Port/interface)

244. Inter prediction (inter prediction or inter frame/picture prediction)

245. Inter prediction block

254. Intra prediction (intra prediction or intra frame/picture prediction)

255. Intra prediction block

260. Mode selection

265. Prediction block (inter prediction block 245 or intra prediction block 255)

FIG. 3

300. Coding system

310. Source device

312. Image source

313 (original) image data

314. Preprocessor/preprocessing unit

315. Preprocessing image data

318. Communication unit/interface

320. Destination device

322. Communication unit/interface

326. Post processor/post processing unit

327. Post-processing image data

328. Display device/unit

330. Transmit/receive/convey (encode) image data fig. 4

410. Current block

420. Reconstruction block

430. Prediction block

440. Prediction error block

FIG. 5

510. Ordering coefficient symbols according to absolute levels of coefficients

520. Symbol estimation

530 CABAC coding

540. Bypass coding

FIG. 6

610. Current block

620. Adjacent pixel columns

630. Adjacent pixel rows

FIG. 7

710. Reconstruction block

720. Prediction block in pixel domain

730. Residual block in pixel domain

740. Neighboring block on left side of current block

750. Difference signal

760. Propagation prediction

770. Adjacent block row

790. Current block

FIG. 8

801. Residual error

802. Context determination

803. Acquiring one or more rows of one or more neighboring blocks

804. Prediction in the pixel domain

805. Transformation

806. Symbol prediction

807. Cost function evaluation

808. Symbol prediction error coding

809. Symbol encoding without prediction

811. Embedding coded symbols and symbol prediction errors in a code stream FIG. 9

122. Symbol prediction

123. Symbol prediction error

970. Entropy coding

971. Encoding image data

FIG. 10

1004. Entropy decoding

209. Reconstructing symbols

FIG. 13

1300. Transformation/quantization

1320. Predicting one or more symbols

1340. Calculating symbol prediction errors

1360. Selecting a context

1380. Entropy coding

FIG. 14

1400. Parsing a code stream

1420. Predicting one or more symbols

1440. Selecting a context

1460. Entropy decoding of symbol prediction errors

1480. Determining one or more symbols

Claims

1. An image block encoding apparatus comprising processing circuitry for:

predicting a symbol of a transform coefficient obtained by transforming the image block signal using one of a plurality of transforms;

Determining a symbol prediction error, the symbol prediction error indicating whether the symbol prediction is correct;

selecting a context for entropy encoding the symbol prediction error according to the one of the plurality of transforms;

encoding the symbol prediction error by applying entropy encoding using the selected context;

the processing circuit is further configured to:

predicting a sign of each of M transform coefficients, M being an integer, greater than 1 and less than a number of transform coefficients obtained by transforming the image block;

determining a symbol prediction error for each of the M transform coefficients;

dividing the M transform coefficients into two lists by a threshold, wherein a first list includes transform coefficients having absolute values greater than the threshold, and a second list includes transform coefficients having absolute values equal to or less than the threshold, comprising: transform coefficients greater than a first threshold are included in a first list, transform coefficients less than a second threshold are included in a second list, and the remaining transform coefficients are included in a third list;

selecting a context for entropy encoding a symbol prediction error of the transform coefficient of the M transform coefficients according to whether the transform coefficient of the M transform coefficients belongs to the first list or the second list or the third list;

The symbol prediction errors of the transform coefficients of the M transform coefficients are encoded by applying entropy encoding using the selected context.

2. The apparatus of claim 1, wherein the processing circuit is further configured to:

the context is selected based on unsigned values of the transform coefficients.

3. The apparatus of claim 1, wherein the processing circuit is further configured to:

the context is selected according to whether the prediction type of the image block is intra prediction or inter prediction.

4. The apparatus of claim 1, wherein the processing circuit is further configured to:

if the prediction type of the image block is intra prediction, the context is selected according to the intra prediction mode used to predict the image block.

5. The apparatus of claim 1, wherein the processing circuit is further configured to:

predicting the sign of the transform coefficient according to a cost function, the cost function comprising estimating discontinuities along boundaries of the image block;

the context is selected according to the number of block edges or the number of neighboring pixels that can be used to estimate a discontinuity on the boundary of the image block.

6. The apparatus of claim 1, wherein the processing circuit is further configured to:

the context is selected according to the ratio of the number of edge pixels available for estimating the discontinuity to the total number of pixels belonging to the image block.

7. The apparatus of claim 1, wherein the processing circuit is further configured to:

the context is selected according to whether the image block is located at the left or upper boundary of the tile where the image block is located.

8. The apparatus of claim 1, wherein the processing circuit is further configured to:

predicting the image block using a DC intra prediction mode or a planar mode;

the context is selected according to whether the image block is predicted using a DC intra prediction mode or a plane mode.

9. The apparatus of claim 1, wherein the processing circuit is further configured to:

the sign of the transform coefficients is predicted according to a cost function, wherein the cost function comprises transform differences between neighboring pixels of a neighboring transformed image block and predictions of the neighboring pixels calculated from a prediction signal of the image block.

10. The apparatus of any of claims 1 to 9, wherein each of the plurality of transforms is associated with one of a plurality of contexts, the processing circuitry further to:

The context is selected from a plurality of contexts such that one of the plurality of transforms is associated with the selected context.

11. The apparatus of claim 10, wherein the processing circuit is further configured to:

selecting a context for entropy encoding the symbol prediction error according to one of the plurality of transforms each associated with one of a plurality of contexts, wherein

If the probabilities of predicted symbol values of a first transform of the plurality of transforms and a second transform of the plurality of transforms differ by less than a predetermined value, the first transform is associated with the same context as the second transform.

12. The apparatus of any one of claims 1 to 9, wherein the processing circuit is further to:

predicting a sign of a transform coefficient obtained by transforming an image block signal using one of a plurality of transforms, wherein

The transformation includes a horizontal transformation and a vertical transformation;

the context is selected according to a combination of the horizontal transform and the vertical transform of the transform.

13. The apparatus of claim 12, wherein each of the plurality of transforms is a combination of a horizontal transform and a vertical transform, the processing circuit further to:

The context is selected such that the combination of the horizontal transform and the vertical transform is associated with the selected context.

14. The apparatus of claim 12, wherein the device comprises a plurality of sensors,

the number of mutually different contexts of the combination of the vertical transform and the horizontal transform is smaller than the number of the combination of the vertical transform and the horizontal transform,

the first combination of vertical transforms and horizontal transforms is associated with the same context as the second combination of vertical transforms and horizontal transforms.

15. The apparatus of any one of claims 1 to 9, wherein the processing circuit is further to: the entropy encoding is performed using CABAC.

16. An apparatus for decoding an image block from a code stream, the apparatus comprising processing circuitry for:

predicting a symbol of a transform coefficient to be inverse transformed using one of a plurality of transforms to obtain an image block signal;

selecting a context for entropy decoding the symbol prediction error according to the one of the plurality of transforms, the symbol prediction error indicating whether the symbol prediction is correct;

decoding the symbol prediction error by applying entropy decoding using a selected context;

Determining the sign of the transform coefficient according to the predicted sign of the transform coefficient and the decoded sign prediction error of the transform coefficient;

the processing circuit is further configured to:

determining a symbol prediction error for each of the M transform coefficients;

17. A method of encoding an image block, comprising the steps of:

selecting a context for entropy encoding the symbol prediction error according to one of the plurality of transforms;

the image block encoding method further includes:

determining a symbol prediction error for each of the M transform coefficients;

18. A method of decoding an image block from a code stream, comprising the steps of:

the method further comprises the steps of:

Determining a symbol prediction error for each of the M transform coefficients;

19. Program code stored on a computer readable medium, characterized in that it comprises instructions which, when executed on a processor, perform all the steps of the method according to claim 17 or 18.